From kimai at Princeton.Edu Sat Aug 16 02:40:42 2014 From: kimai at Princeton.Edu (Kosuke Imai) Date: Fri, 15 Aug 2014 20:40:42 -0400 Subject: [Mediation-information] Sensitivity analysis In-Reply-To: <3625E4151D36B04DA27619B242B6EA0D1B54362D@MBX01.uva.nl> References: <3625E4151D36B04DA27619B242B6EA0D1B54362D@MBX01.uva.nl> Message-ID: Alina, For the examples of sensitivity analyses, you should look at our APSR paper. Our psychological methods paper also has some examples too. For the post-treatment variable issue, here is the paper you can look at: http://imai.princeton.edu/research/medsens.html The procedure is implemented in the R package too: http://imai.princeton.edu/research/files/mediationR2.pdf Best, Kosuke Kosuke Imai Department of Politics Princeton University http://imai.princeton.edu On Fri, Aug 15, 2014 at 9:51 AM, Feinholdt, A. wrote: > Dear Kosuke, > > a couple of days ago, I was in contact with you concerning a STATA code > for the sensitivity analysis. Fortunately, I was able to understand and run > the R code. However, I am currently facing two problems: Since your > analysis is still new, there are only a handful of studies which use the > sensitivity analysis to strengthen their results. However, from these > studies it is unclear what and how to report findings from the sensitivity > analysis. Therefore, I would like to seek your advice on this matter. > That is, what do you think is essential to be reported from a sensitivity > analysis? > > The other problem concerns my findings: The goal of my study (an > experimental online survey in which participants were randomly assigned to > one of four conditions) was to show that framing effects would be serially > mediated through two emotions. To support the validity of a serial > mediation, I have first run a "normal" causal mediation analysis with R > while ignoring the possibility of a potential post-treatment confounder. > Here the findings supported the notion of an ACME through one of the two > emotions. Next, I have tested the causal mediation analysis with > confounding by an alternative pathway. This analysis demonstrated that ACME > was no longer significant - neither in the treatment nor in the control > group. As such, I have concluded that the initial assumption was no longer > tenable once the post-treatment confounder was accounted for. In addition, > I have argued that under conditions like these it would be important to > refer to other analytical strategies such as the serial mediation analysis > - since this one does not presupposes causal independence between pathways. > I can assume that this description is pretty vague but it may be enough > for you to conclude, whether the sensitivity analysis can be used to > support the application of a serial mediation analysis or not? > > I can imagine that you have a busy schedule but maybe you can find some > time to consult me on the aforementioned issues. > Thanks in advance for reading my e-mail and I hope to hear from you. > > Best wishes, > Alina Feinholdt > *PhD candidate* > *Amsterdam School of Communication Research* > > > > > ------------------------------ > *Von:* Kosuke Imai [kimai at princeton.edu] > *Gesendet:* Mittwoch, 6. August 2014 11:34 > *An:* Feinholdt, A. > *Betreff:* Re: STATA - sensitivity analysis > > Dear Alina, > > Unfortunately, we don't have this code in STATA. But, you can use our > R package, which is pretty straightforward even if you don't know much > about R. Try this tutorial, which gives you a step-by-step instruction: > http://imai.princeton.edu/research/files/mediationR2.pdf > > Best, > Kosuke > > --------------------------------------------------------- > Kosuke Imai Office: Corwin Hall 036 > Professor Phone: 609-258-6601 > Department of Politics Fax: 609-258-1110 > Princeton University Email: kimai at Princeton.Edu > Princeton, NJ 08544-1012 http://imai.princeton.edu > --------------------------------------------------------- > > > On Wed, Aug 6, 2014 at 3:52 AM, Feinholdt, A. wrote: > >> Dear prof. Imai, >> >> my name is Alina Feinholdt and I am a PhD student at the University of >> Amsterdam, the Netherlands. I just recently read one of your papers on >> "Identification and Sensitivity Analysis for Multiple Causal Mechanisms: >> Revisiting Evidence from Framing Experiment". Since I am dealing myself >> with potential post-treatment confounders (two emotions are studied as >> pathways following news framing), I would like to run both the mediation >> effect analysis and the sensitivity analysis with STATA. However, so far I >> could only find an R code dealing with causally dependent multiple >> mechanisms. My question is now if you also have a STATA code that would >> allow me to run an analysis for causally dependent multiple mechanisms? If >> not is there another way of running the same analysis with the STATA? That >> is, which calculations would I need to do in order to analyse my data for >> potential post-treatment confounders? >> >> I would much appreciate your response as I currently do not know who >> else could help me with it. >> >> Best wishes, >> Alina Feinholdt >> *PhD candidate* >> *Amsterdam School of Communication Research* >> >> *Universiteit van Amsterdam* >> *Kloveniersburgwal 48,* >> *1012 CX Amsterdam, * >> *The Netherlands* >> >> > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From angel.rodriguez at matiainstituto.net Tue Aug 26 00:14:31 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Tue, 26 Aug 2014 00:14:31 +0200 Subject: [Mediation-information] Causally dependent mediators Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B21B@MATIAEXCH.matiaf.local> Dear subscribers, I'm interested in using mediation for the following conceptual model applied to survey data: infant deprivation (binary) --> education level (ordered categorical) --> economic position (continous) --> aging well (dichotomous outcome) I assume that, apart from the mediation effects, each link has direct effects on the outcome. I've read in "Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies" (Imai et al 2011), that studying education as the main mediator wouldn't be a problem, because the sequential ignorability assumption would be met due to the fact that the second mediator is downstream (figure 6d of that paper). That would be my first model. What should I do to know if the effect of education is partly or completely mediated by economic position? I understand I cannot use economic position as a main mediator in the model deprivation --> economic position --> aging well, because there would be an upstream mediator (education), which violates the sequential ignorability. I should use multimed then (my second model): R> Xnames <- c("age","gender", "region") R> set.seed(2014) R> m.med <- multimed(outcome = "agingwell", med.main = "economic", med.alt = "education", + treat = "deprivation", covariates = Xnames, + data = framing, sims = 1000) R> summary(m.med) If I get a large mediating effect of "education" in my first model and a negligible effect of "economic" in the second, could I conclude that all the mediating effect of "education" on the relationship between "deprivation" and aging well is driven by "economic"? I suppose I cannot analyse multilevel data with multimed. Thank you very much. Angel Rodriguez-Laso Research Project Manager Matia Instituto Gerontol?gico -------------- next part -------------- An HTML attachment was scrubbed... URL: From senor_massao at hotmail.com Tue Aug 26 15:01:06 2014 From: senor_massao at hotmail.com (Mashhood Sheikh) Date: Tue, 26 Aug 2014 14:01:06 +0100 Subject: [Mediation-information] Mediation-information Digest, Vol 25, Issue 2 In-Reply-To: References: Message-ID: In order to assess if the effect of education on ageing well is partly or completely mediated by economic position, you can run this model (adjusted for all potential confounders): education level --> economic position --> aging well? --- Mashhood Ahmed Sheikh Research Fellow, Department of Community Medicine,? Faculty of Health, University of Troms?, Norway > From: mediation-information-request at lists.r-forge.r-project.org > Subject: Mediation-information Digest, Vol 25, Issue 2 > To: mediation-information at lists.r-forge.r-project.org > Date: Tue, 26 Aug 2014 12:00:22 +0200 > > Send Mediation-information mailing list submissions to > mediation-information at lists.r-forge.r-project.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > or, via email, send a message with subject or body 'help' to > mediation-information-request at lists.r-forge.r-project.org > > You can reach the person managing the list at > mediation-information-owner at lists.r-forge.r-project.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Mediation-information digest..." > > > Today's Topics: > > 1. Causally dependent mediators (Angel Rodriguez) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 26 Aug 2014 00:14:31 +0200 > From: "Angel Rodriguez" > To: > Subject: [Mediation-information] Causally dependent mediators > Message-ID: > <8564BCD7D26E0D40872F1A132C8BBB250258B21B at MATIAEXCH.matiaf.local> > Content-Type: text/plain; charset="iso-8859-1" > > > Dear subscribers, > > > I'm interested in using mediation for the following conceptual model applied to survey data: > > infant deprivation (binary) --> education level (ordered categorical) --> economic position (continous) --> aging well (dichotomous outcome) > > I assume that, apart from the mediation effects, each link has direct effects on the outcome. > > I've read in "Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies" (Imai et al 2011), that studying education as the main mediator wouldn't be a problem, because the sequential ignorability assumption would be met due to the fact that the second mediator is downstream (figure 6d of that paper). That would be my first model. > > What should I do to know if the effect of education is partly or completely mediated by economic position? I understand I cannot use economic position as a main mediator in the model deprivation --> economic position --> aging well, because there would be an upstream mediator (education), which violates the sequential ignorability. I should use multimed then (my second model): > > > R> Xnames <- c("age","gender", "region") > R> set.seed(2014) > R> m.med <- multimed(outcome = "agingwell", med.main = "economic", med.alt = "education", > + treat = "deprivation", covariates = Xnames, > + data = framing, sims = 1000) > R> summary(m.med) > > > If I get a large mediating effect of "education" in my first model and a negligible effect of "economic" in the second, could I conclude that all the mediating effect of "education" on the relationship between "deprivation" and aging well is driven by "economic"? > > I suppose I cannot analyse multilevel data with multimed. > > > > Thank you very much. > > Angel Rodriguez-Laso > Research Project Manager > Matia Instituto Gerontol?gico > > > > > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > > ------------------------------ > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > End of Mediation-information Digest, Vol 25, Issue 2 > **************************************************** From angel.rodriguez at matiainstituto.net Tue Aug 26 18:24:37 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Tue, 26 Aug 2014 18:24:37 +0200 Subject: [Mediation-information] Causally dependent mediators References: Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B225@MATIAEXCH.matiaf.local> Thank you, Mashhood, for your answer. If I understand correctly, you recommend I run two models: infant deprivation --> education level --> aging well (plus age, gender, region) education level --> economic position --> aging well (plus age, gender, region) Because the effects of infant deprivation are considered in the first model, there is no need to include it as a confounder in the second model. Is that correct? Angel Rodriguez-Laso -----Mensaje original----- De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Mashhood Sheikh Enviado el: mar 26/08/2014 15:01 Para: mediation-information at lists.r-forge.r-project.org Asunto: Re: [Mediation-information] Mediation-information Digest, Vol 25,Issue 2 In order to assess if the effect of education on ageing well is partly or completely mediated by economic position, you can run this model (adjusted for all potential confounders): education level --> economic position --> aging well? --- Mashhood Ahmed Sheikh Research Fellow, Department of Community Medicine,? Faculty of Health, University of Troms?, Norway > From: mediation-information-request at lists.r-forge.r-project.org > Subject: Mediation-information Digest, Vol 25, Issue 2 > To: mediation-information at lists.r-forge.r-project.org > Date: Tue, 26 Aug 2014 12:00:22 +0200 > > Send Mediation-information mailing list submissions to > mediation-information at lists.r-forge.r-project.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > or, via email, send a message with subject or body 'help' to > mediation-information-request at lists.r-forge.r-project.org > > You can reach the person managing the list at > mediation-information-owner at lists.r-forge.r-project.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Mediation-information digest..." > > > Today's Topics: > > 1. Causally dependent mediators (Angel Rodriguez) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 26 Aug 2014 00:14:31 +0200 > From: "Angel Rodriguez" > To: > Subject: [Mediation-information] Causally dependent mediators > Message-ID: > <8564BCD7D26E0D40872F1A132C8BBB250258B21B at MATIAEXCH.matiaf.local> > Content-Type: text/plain; charset="iso-8859-1" > > > Dear subscribers, > > > I'm interested in using mediation for the following conceptual model applied to survey data: > > infant deprivation (binary) --> education level (ordered categorical) --> economic position (continous) --> aging well (dichotomous outcome) > > I assume that, apart from the mediation effects, each link has direct effects on the outcome. > > I've read in "Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies" (Imai et al 2011), that studying education as the main mediator wouldn't be a problem, because the sequential ignorability assumption would be met due to the fact that the second mediator is downstream (figure 6d of that paper). That would be my first model. > > What should I do to know if the effect of education is partly or completely mediated by economic position? I understand I cannot use economic position as a main mediator in the model deprivation --> economic position --> aging well, because there would be an upstream mediator (education), which violates the sequential ignorability. I should use multimed then (my second model): > > > R> Xnames <- c("age","gender", "region") > R> set.seed(2014) > R> m.med <- multimed(outcome = "agingwell", med.main = "economic", med.alt = "education", > + treat = "deprivation", covariates = Xnames, > + data = framing, sims = 1000) > R> summary(m.med) > > > If I get a large mediating effect of "education" in my first model and a negligible effect of "economic" in the second, could I conclude that all the mediating effect of "education" on the relationship between "deprivation" and aging well is driven by "economic"? > > I suppose I cannot analyse multilevel data with multimed. > > > > Thank you very much. > > Angel Rodriguez-Laso > Research Project Manager > Matia Instituto Gerontol?gico > > > > > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > > ------------------------------ > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > End of Mediation-information Digest, Vol 25, Issue 2 > **************************************************** _______________________________________________ Mediation-information mailing list Mediation-information at lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -------------- next part -------------- An HTML attachment was scrubbed... URL: From angel.rodriguez at matiainstituto.net Tue Aug 26 18:49:49 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Tue, 26 Aug 2014 18:49:49 +0200 Subject: [Mediation-information] Multilevel analysis for complex surveys Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B226@MATIAEXCH.matiaf.local> Dear subscribers, I want to carry out a mediation analysis with variables collected in a survey (for more information, see the thread "causally dependent mediators"). It was a complex survey with strata and individuals clustered in census sections. I do not have census section variables, but I want the models to be aware that individuals are not independent because they're grouped in census sections. My first question is if I should use mediation with glmer. In that case, my second question would be about the syntax to carry out this. In the example in Tingley's "mediation: R Package for Causal Mediation Analysis", the proposed models are: R> library(lme4) R> set.seed(2014) R> med.fit <- glmer(attachment ~ catholic + gender + income + pared + (1|SCH_ID), + family = binomial(link = "logit"), data = student) R> out.fit <- glmer(fight ~ catholic*attachment + + gender + income + pared + (1 + attachment|SCH_ID), + family = binomial(link = "logit"), data = student) I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. Thank you very much, Angel Rodriguez-Laso Research Project Manager Matia Instituto Gerontologico -------------- next part -------------- An HTML attachment was scrubbed... URL: From hirose at Princeton.EDU Tue Aug 26 19:02:52 2014 From: hirose at Princeton.EDU (Kentaro Hirose) Date: Tue, 26 Aug 2014 17:02:52 +0000 Subject: [Mediation-information] Multilevel analysis for complex surveys In-Reply-To: <8564BCD7D26E0D40872F1A132C8BBB250258B226@MATIAEXCH.matiaf.local> References: <8564BCD7D26E0D40872F1A132C8BBB250258B226@MATIAEXCH.matiaf.local> Message-ID: <7F71CA3EDBFDB742900C47C9596B90233909ADDF@CSGMBX201W.pu.win.princeton.edu> Hi Angel, > I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. This is just fitting a varying-intercept model for the mediator and a varying-intercept-and-slope model for the outcome. You could also fit two varying-intercept models for both the mediator and outcome. And similarly, you could also fit two varying-intercept-and-slope models for both the mediator and outcome. It's all up to your theory. > My first question is if I should use mediation with glmer. If your dependent variable is binary and you want to use a multilevel model, then you should use glmer. Best, Kentaro Hirose Postdoctoral Fellow Department of Politics Princeton University ________________________________________ From: mediation-information-bounces at r-forge.wu-wien.ac.at [mediation-information-bounces at r-forge.wu-wien.ac.at] on behalf of Angel Rodriguez [angel.rodriguez at matiainstituto.net] Sent: Tuesday, August 26, 2014 12:49 PM To: mediation-information at r-forge.wu-wien.ac.at Subject: [Mediation-information] Multilevel analysis for complex surveys Dear subscribers, I want to carry out a mediation analysis with variables collected in a survey (for more information, see the thread "causally dependent mediators"). It was a complex survey with strata and individuals clustered in census sections. I do not have census section variables, but I want the models to be aware that individuals are not independent because they're grouped in census sections. My first question is if I should use mediation with glmer. In that case, my second question would be about the syntax to carry out this. In the example in Tingley's "mediation: R Package for Causal Mediation Analysis", the proposed models are: R> library(lme4) R> set.seed(2014) R> med.fit <- glmer(attachment ~ catholic + gender + income + pared + (1|SCH_ID), + family = binomial(link = "logit"), data = student) R> out.fit <- glmer(fight ~ catholic*attachment + + gender + income + pared + (1 + attachment|SCH_ID), + family = binomial(link = "logit"), data = student) I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. Thank you very much, Angel Rodriguez-Laso Research Project Manager Matia Instituto Gerontologico From angel.rodriguez at matiainstituto.net Tue Aug 26 19:36:15 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Tue, 26 Aug 2014 19:36:15 +0200 Subject: [Mediation-information] Multilevel analysis for complex surveys References: <8564BCD7D26E0D40872F1A132C8BBB250258B226@MATIAEXCH.matiaf.local> <7F71CA3EDBFDB742900C47C9596B90233909ADDF@CSGMBX201W.pu.win.princeton.edu> Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B229@MATIAEXCH.matiaf.local> Thank you very much, Kentaro. Angel -----Mensaje original----- De: Kentaro Hirose [mailto:hirose at Princeton.EDU] Enviado el: mar 26/08/2014 19:02 Para: Angel Rodriguez; mediation-information at r-forge.wu-wien.ac.at Asunto: RE: Multilevel analysis for complex surveys Hi Angel, > I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. This is just fitting a varying-intercept model for the mediator and a varying-intercept-and-slope model for the outcome. You could also fit two varying-intercept models for both the mediator and outcome. And similarly, you could also fit two varying-intercept-and-slope models for both the mediator and outcome. It's all up to your theory. > My first question is if I should use mediation with glmer. If your dependent variable is binary and you want to use a multilevel model, then you should use glmer. Best, Kentaro Hirose Postdoctoral Fellow Department of Politics Princeton University ________________________________________ From: mediation-information-bounces at r-forge.wu-wien.ac.at [mediation-information-bounces at r-forge.wu-wien.ac.at] on behalf of Angel Rodriguez [angel.rodriguez at matiainstituto.net] Sent: Tuesday, August 26, 2014 12:49 PM To: mediation-information at r-forge.wu-wien.ac.at Subject: [Mediation-information] Multilevel analysis for complex surveys Dear subscribers, I want to carry out a mediation analysis with variables collected in a survey (for more information, see the thread "causally dependent mediators"). It was a complex survey with strata and individuals clustered in census sections. I do not have census section variables, but I want the models to be aware that individuals are not independent because they're grouped in census sections. My first question is if I should use mediation with glmer. In that case, my second question would be about the syntax to carry out this. In the example in Tingley's "mediation: R Package for Causal Mediation Analysis", the proposed models are: R> library(lme4) R> set.seed(2014) R> med.fit <- glmer(attachment ~ catholic + gender + income + pared + (1|SCH_ID), + family = binomial(link = "logit"), data = student) R> out.fit <- glmer(fight ~ catholic*attachment + + gender + income + pared + (1 + attachment|SCH_ID), + family = binomial(link = "logit"), data = student) I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. Thank you very much, Angel Rodriguez-Laso Research Project Manager Matia Instituto Gerontologico -------------- next part -------------- An HTML attachment was scrubbed... URL: From senor_massao at hotmail.com Tue Aug 26 20:56:12 2014 From: senor_massao at hotmail.com (Mashhood Sheikh) Date: Tue, 26 Aug 2014 19:56:12 +0100 Subject: [Mediation-information] Mediation-information Digest, Vol 25, Issue 3 In-Reply-To: References: Message-ID: Hi, In the model (education level --> economic position --> aging well) you should include all the variables that potentially confound the association between education level and economic position, education level and ageing well, and economic position and ageing well.? Best wishes, Mashhood --- Mashhood Ahmed Sheikh Research Fellow, Department of Community Medicine Faculty of Health, University of Troms?, Norway ---------------------------------------- > From: mediation-information-request at lists.r-forge.r-project.org > Subject: Mediation-information Digest, Vol 25, Issue 3 > To: mediation-information at lists.r-forge.r-project.org > Date: Tue, 26 Aug 2014 19:03:02 +0200 > > Send Mediation-information mailing list submissions to > mediation-information at lists.r-forge.r-project.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > or, via email, send a message with subject or body 'help' to > mediation-information-request at lists.r-forge.r-project.org > > You can reach the person managing the list at > mediation-information-owner at lists.r-forge.r-project.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Mediation-information digest..." > > > Today's Topics: > > 1. Re: Mediation-information Digest, Vol 25, Issue 2 > (Mashhood Sheikh) > 2. Causally dependent mediators (Angel Rodriguez) > 3. Multilevel analysis for complex surveys (Angel Rodriguez) > 4. Re: Multilevel analysis for complex surveys (Kentaro Hirose) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 26 Aug 2014 14:01:06 +0100 > From: Mashhood Sheikh > To: "mediation-information at lists.r-forge.r-project.org" > > Subject: Re: [Mediation-information] Mediation-information Digest, Vol > 25, Issue 2 > Message-ID: > Content-Type: text/plain; charset="iso-8859-1" > > In order to assess if the effect of education on ageing well is partly or completely mediated by economic position, you can run this model (adjusted for all potential confounders): > > education level --> economic position --> aging well? > > --- > Mashhood Ahmed Sheikh > Research Fellow, > Department of Community Medicine,? > Faculty of Health, > University of Troms?, > Norway > > > >> From: mediation-information-request at lists.r-forge.r-project.org >> Subject: Mediation-information Digest, Vol 25, Issue 2 >> To: mediation-information at lists.r-forge.r-project.org >> Date: Tue, 26 Aug 2014 12:00:22 +0200 >> >> Send Mediation-information mailing list submissions to >> mediation-information at lists.r-forge.r-project.org >> >> To subscribe or unsubscribe via the World Wide Web, visit >> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information >> >> or, via email, send a message with subject or body 'help' to >> mediation-information-request at lists.r-forge.r-project.org >> >> You can reach the person managing the list at >> mediation-information-owner at lists.r-forge.r-project.org >> >> When replying, please edit your Subject line so it is more specific >> than "Re: Contents of Mediation-information digest..." >> >> >> Today's Topics: >> >> 1. Causally dependent mediators (Angel Rodriguez) >> >> >> ---------------------------------------------------------------------- >> >> Message: 1 >> Date: Tue, 26 Aug 2014 00:14:31 +0200 >> From: "Angel Rodriguez" >> To: >> Subject: [Mediation-information] Causally dependent mediators >> Message-ID: >> <8564BCD7D26E0D40872F1A132C8BBB250258B21B at MATIAEXCH.matiaf.local> >> Content-Type: text/plain; charset="iso-8859-1" >> >> >> Dear subscribers, >> >> >> I'm interested in using mediation for the following conceptual model applied to survey data: >> >> infant deprivation (binary) --> education level (ordered categorical) --> economic position (continous) --> aging well (dichotomous outcome) >> >> I assume that, apart from the mediation effects, each link has direct effects on the outcome. >> >> I've read in "Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies" (Imai et al 2011), that studying education as the main mediator wouldn't be a problem, because the sequential ignorability assumption would be met due to the fact that the second mediator is downstream (figure 6d of that paper). That would be my first model. >> >> What should I do to know if the effect of education is partly or completely mediated by economic position? I understand I cannot use economic position as a main mediator in the model deprivation --> economic position --> aging well, because there would be an upstream mediator (education), which violates the sequential ignorability. I should use multimed then (my second model): >> >> >> R> Xnames <- c("age","gender", "region") >> R> set.seed(2014) >> R> m.med <- multimed(outcome = "agingwell", med.main = "economic", med.alt = "education", >> + treat = "deprivation", covariates = Xnames, >> + data = framing, sims = 1000) >> R> summary(m.med) >> >> >> If I get a large mediating effect of "education" in my first model and a negligible effect of "economic" in the second, could I conclude that all the mediating effect of "education" on the relationship between "deprivation" and aging well is driven by "economic"? >> >> I suppose I cannot analyse multilevel data with multimed. >> >> >> >> Thank you very much. >> >> Angel Rodriguez-Laso >> Research Project Manager >> Matia Instituto Gerontol?gico >> >> >> >> >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: >> >> ------------------------------ >> >> _______________________________________________ >> Mediation-information mailing list >> Mediation-information at lists.r-forge.r-project.org >> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information >> >> End of Mediation-information Digest, Vol 25, Issue 2 >> **************************************************** > > > ------------------------------ > > Message: 2 > Date: Tue, 26 Aug 2014 18:24:37 +0200 > From: "Angel Rodriguez" > To: > Subject: [Mediation-information] Causally dependent mediators > Message-ID: > <8564BCD7D26E0D40872F1A132C8BBB250258B225 at MATIAEXCH.matiaf.local> > Content-Type: text/plain; charset="iso-8859-1" > > Thank you, Mashhood, for your answer. > > If I understand correctly, you recommend I run two models: > > infant deprivation --> education level --> aging well (plus age, gender, region) > > education level --> economic position --> aging well (plus age, gender, region) > > > Because the effects of infant deprivation are considered in the first model, there is no need to include it as a confounder in the second model. > > Is that correct? > > Angel Rodriguez-Laso > > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Mashhood Sheikh > Enviado el: mar 26/08/2014 15:01 > Para: mediation-information at lists.r-forge.r-project.org > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol 25,Issue 2 > > In order to assess if the effect of education on ageing well is partly or completely mediated by economic position, you can run this model (adjusted for all potential confounders): > > education level --> economic position --> aging well? > > --- > Mashhood Ahmed Sheikh > Research Fellow, > Department of Community Medicine,? > Faculty of Health, > University of Troms?, > Norway > > > >> From: mediation-information-request at lists.r-forge.r-project.org >> Subject: Mediation-information Digest, Vol 25, Issue 2 >> To: mediation-information at lists.r-forge.r-project.org >> Date: Tue, 26 Aug 2014 12:00:22 +0200 >> >> Send Mediation-information mailing list submissions to >> mediation-information at lists.r-forge.r-project.org >> >> To subscribe or unsubscribe via the World Wide Web, visit >> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information >> >> or, via email, send a message with subject or body 'help' to >> mediation-information-request at lists.r-forge.r-project.org >> >> You can reach the person managing the list at >> mediation-information-owner at lists.r-forge.r-project.org >> >> When replying, please edit your Subject line so it is more specific >> than "Re: Contents of Mediation-information digest..." >> >> >> Today's Topics: >> >> 1. Causally dependent mediators (Angel Rodriguez) >> >> >> ---------------------------------------------------------------------- >> >> Message: 1 >> Date: Tue, 26 Aug 2014 00:14:31 +0200 >> From: "Angel Rodriguez" >> To: >> Subject: [Mediation-information] Causally dependent mediators >> Message-ID: >> <8564BCD7D26E0D40872F1A132C8BBB250258B21B at MATIAEXCH.matiaf.local> >> Content-Type: text/plain; charset="iso-8859-1" >> >> >> Dear subscribers, >> >> >> I'm interested in using mediation for the following conceptual model applied to survey data: >> >> infant deprivation (binary) --> education level (ordered categorical) --> economic position (continous) --> aging well (dichotomous outcome) >> >> I assume that, apart from the mediation effects, each link has direct effects on the outcome. >> >> I've read in "Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies" (Imai et al 2011), that studying education as the main mediator wouldn't be a problem, because the sequential ignorability assumption would be met due to the fact that the second mediator is downstream (figure 6d of that paper). That would be my first model. >> >> What should I do to know if the effect of education is partly or completely mediated by economic position? I understand I cannot use economic position as a main mediator in the model deprivation --> economic position --> aging well, because there would be an upstream mediator (education), which violates the sequential ignorability. I should use multimed then (my second model): >> >> >> R> Xnames <- c("age","gender", "region") >> R> set.seed(2014) >> R> m.med <- multimed(outcome = "agingwell", med.main = "economic", med.alt = "education", >> + treat = "deprivation", covariates = Xnames, >> + data = framing, sims = 1000) >> R> summary(m.med) >> >> >> If I get a large mediating effect of "education" in my first model and a negligible effect of "economic" in the second, could I conclude that all the mediating effect of "education" on the relationship between "deprivation" and aging well is driven by "economic"? >> >> I suppose I cannot analyse multilevel data with multimed. >> >> >> >> Thank you very much. >> >> Angel Rodriguez-Laso >> Research Project Manager >> Matia Instituto Gerontol?gico >> >> >> >> >> -------------- next part -------------- >> An HTML attachment was scrubbed... >> URL: >> >> ------------------------------ >> >> _______________________________________________ >> Mediation-information mailing list >> Mediation-information at lists.r-forge.r-project.org >> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information >> >> End of Mediation-information Digest, Vol 25, Issue 2 >> **************************************************** > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > > ------------------------------ > > Message: 3 > Date: Tue, 26 Aug 2014 18:49:49 +0200 > From: "Angel Rodriguez" > To: > Subject: [Mediation-information] Multilevel analysis for complex > surveys > Message-ID: > <8564BCD7D26E0D40872F1A132C8BBB250258B226 at MATIAEXCH.matiaf.local> > Content-Type: text/plain; charset="iso-8859-1" > > Dear subscribers, > > I want to carry out a mediation analysis with variables collected in a survey (for more information, see the thread "causally dependent mediators"). It was a complex survey with strata and individuals clustered in census sections. I do not have census section variables, but I want the models to be aware that individuals are not independent because they're grouped in census sections. > > My first question is if I should use mediation with glmer. > > In that case, my second question would be about the syntax to carry out this. In the example in Tingley's "mediation: R Package for Causal Mediation Analysis", the proposed models are: > > > R> library(lme4) > R> set.seed(2014) > R> med.fit <- glmer(attachment ~ catholic + gender + income + pared + (1|SCH_ID), > + family = binomial(link = "logit"), data = student) > R> out.fit <- glmer(fight ~ catholic*attachment + > + gender + income + pared + (1 + attachment|SCH_ID), > + family = binomial(link = "logit"), data = student) > > I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. > > Thank you very much, > > Angel Rodriguez-Laso > Research Project Manager > Matia Instituto Gerontologico > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > > ------------------------------ > > Message: 4 > Date: Tue, 26 Aug 2014 17:02:52 +0000 > From: Kentaro Hirose > To: Angel Rodriguez , > "mediation-information at r-forge.wu-wien.ac.at" > > Subject: Re: [Mediation-information] Multilevel analysis for complex > surveys > Message-ID: > <7F71CA3EDBFDB742900C47C9596B90233909ADDF at CSGMBX201W.pu.win.princeton.edu> > > Content-Type: text/plain; charset="us-ascii" > > Hi Angel, > >> I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. > > This is just fitting a varying-intercept model for the mediator and a varying-intercept-and-slope model for the outcome. You could also fit two varying-intercept models for both the mediator and outcome. And similarly, you could also fit two varying-intercept-and-slope models for both the mediator and outcome. It's all up to your theory. > > >> My first question is if I should use mediation with glmer. > > If your dependent variable is binary and you want to use a multilevel model, then you should use glmer. > > > Best, > Kentaro Hirose > > Postdoctoral Fellow > Department of Politics > Princeton University > > > > ________________________________________ > From: mediation-information-bounces at r-forge.wu-wien.ac.at [mediation-information-bounces at r-forge.wu-wien.ac.at] on behalf of Angel Rodriguez [angel.rodriguez at matiainstituto.net] > Sent: Tuesday, August 26, 2014 12:49 PM > To: mediation-information at r-forge.wu-wien.ac.at > Subject: [Mediation-information] Multilevel analysis for complex surveys > > Dear subscribers, > > I want to carry out a mediation analysis with variables collected in a survey (for more information, see the thread "causally dependent mediators"). It was a complex survey with strata and individuals clustered in census sections. I do not have census section variables, but I want the models to be aware that individuals are not independent because they're grouped in census sections. > > My first question is if I should use mediation with glmer. > > In that case, my second question would be about the syntax to carry out this. In the example in Tingley's "mediation: R Package for Causal Mediation Analysis", the proposed models are: > > > R> library(lme4) > R> set.seed(2014) > R> med.fit <- glmer(attachment ~ catholic + gender + income + pared + (1|SCH_ID), > + family = binomial(link = "logit"), data = student) > R> out.fit <- glmer(fight ~ catholic*attachment + > + gender + income + pared + (1 + attachment|SCH_ID), > + family = binomial(link = "logit"), data = student) > > I don't understand why in med.fit the term related to the clustering units is (1|SCH_ID), while (1 + attachment|SCH_ID) is the term in the out.fit. > > Thank you very much, > > Angel Rodriguez-Laso > Research Project Manager > Matia Instituto Gerontologico > > > ------------------------------ > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > End of Mediation-information Digest, Vol 25, Issue 3 > **************************************************** From angel.rodriguez at matiainstituto.net Wed Aug 27 11:38:53 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Wed, 27 Aug 2014 11:38:53 +0200 Subject: [Mediation-information] Mediation-information Digest, Vol 25, Issue 3 References: Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> Thank you again, Mashhood. What I do not understand yet is why we need a package like multimed if we can decompose a model with two causally dependent mediators (infant deprivation --> education level --> economic position --> aging well) into two models (infant deprivation --> education level --> aging well & education level --> economic position --> aging well). Following your advise, I should include infant deprivation as a confounder in the model education level --> economic position --> aging well. But if deprivation conditions education strongly, I could end not finding any direct nor indirect effect of education on aging well. Best regards, Angel Rodriguez-Laso -----Mensaje original----- De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Mashhood Sheikh Enviado el: mar 26/08/2014 20:56 Para: mediation-information at lists.r-forge.r-project.org Asunto: Re: [Mediation-information] Mediation-information Digest, Vol 25,Issue 3 Hi, In the model (education level --> economic position --> aging well) you should include all the variables that potentially confound the association between education level and economic position, education level and ageing well, and economic position and ageing well.? Best wishes, Mashhood --- Mashhood Ahmed Sheikh Research Fellow, Department of Community Medicine Faculty of Health, University of Troms?, Norway -------------- next part -------------- An HTML attachment was scrubbed... URL: From teppei at mit.edu Wed Aug 27 15:27:03 2014 From: teppei at mit.edu (Teppei Yamamoto) Date: Wed, 27 Aug 2014 09:27:03 -0400 Subject: [Mediation-information] Mediation-information Digest, Vol 25, Issue 3 In-Reply-To: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> References: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> Message-ID: <53FDDCA7.8060203@mit.edu> Hi Angel, The difference between the two estimates (multimed vs. mediation analysis on education level as the "treatment") is that the former is about the effect of manipulating infant deprivation on aging whereas the latter is about the effect of manipulating education directly. That is, they are estimating effects of different kinds of hypothetical interventions and how they are transmitted through the variables that are assumed to be causally descendant. So I think your choice boils down to what causal quantity you are trying to estimate. Best, Teppei (8/27/14, 5:38 AM), Angel Rodriguez wrote: > Thank you again, Mashhood. > > What I do not understand yet is why we need a package like multimed if > we can decompose a model with two causally dependent mediators (infant > deprivation --> education level --> economic position --> aging well) > into two models (infant deprivation --> education level --> aging well & > education level --> economic position --> aging well). > > Following your advise, I should include infant deprivation as a > confounder in the model education level --> economic position --> aging > well. But if deprivation conditions education strongly, I could end not > finding any direct nor indirect effect of education on aging well. > > Best regards, > > Angel Rodriguez-Laso > > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > de Mashhood Sheikh > Enviado el: mar 26/08/2014 20:56 > Para: mediation-information at lists.r-forge.r-project.org > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > 25,Issue 3 > > > Hi, > > In the model (education level --> economic position --> aging well) you > should include all the variables that potentially confound the > association between education level and economic position, education > level and ageing well, and economic position and ageing well. > > Best wishes, > Mashhood > --- > Mashhood Ahmed Sheikh > Research Fellow, > Department of Community Medicine > Faculty of Health, > University of Troms?, > Norway > > > > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -- ==================================== Teppei Yamamoto Assistant Professor Department of Political Science Massachusetts Institute of Technology http://web.mit.edu/teppei/www/ ==================================== From angel.rodriguez at matiainstituto.net Wed Aug 27 16:35:13 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Wed, 27 Aug 2014 16:35:13 +0200 Subject: [Mediation-information] Causally dependent mediators References: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> <53FDDCA7.8060203@mit.edu> Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B234@MATIAEXCH.matiaf.local> Thank you for your answers, Teppei. So, would these two models make sense? 1) deprivation --> education level --> aging well (covariates: age, gender, region) (with R mediation package, to study the direct and indirect effects of deprivation on aging well) 2) education level --> economic position --> aging well (covariates: DEPRIVATION, age, gender, region) (with R mediation package, to study the direct and indirect effecs on education on aging well, net of deprivation) Best, Angel -----Mensaje original----- De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Teppei Yamamoto Enviado el: mi? 27/08/2014 15:27 Para: Mediation-information at r-forge.wu-wien.ac.at Asunto: Re: [Mediation-information] Mediation-information Digest, Vol 25, Issue 3 Hi Angel, The difference between the two estimates (multimed vs. mediation analysis on education level as the "treatment") is that the former is about the effect of manipulating infant deprivation on aging whereas the latter is about the effect of manipulating education directly. That is, they are estimating effects of different kinds of hypothetical interventions and how they are transmitted through the variables that are assumed to be causally descendant. So I think your choice boils down to what causal quantity you are trying to estimate. Best, Teppei (8/27/14, 5:38 AM), Angel Rodriguez wrote: > Thank you again, Mashhood. > > What I do not understand yet is why we need a package like multimed if > we can decompose a model with two causally dependent mediators (infant > deprivation --> education level --> economic position --> aging well) > into two models (infant deprivation --> education level --> aging well & > education level --> economic position --> aging well). > > Following your advise, I should include infant deprivation as a > confounder in the model education level --> economic position --> aging > well. But if deprivation conditions education strongly, I could end not > finding any direct nor indirect effect of education on aging well. > > Best regards, > > Angel Rodriguez-Laso > > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > de Mashhood Sheikh > Enviado el: mar 26/08/2014 20:56 > Para: mediation-information at lists.r-forge.r-project.org > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > 25,Issue 3 > > > Hi, > > In the model (education level --> economic position --> aging well) you > should include all the variables that potentially confound the > association between education level and economic position, education > level and ageing well, and economic position and ageing well. > > Best wishes, > Mashhood > --- > Mashhood Ahmed Sheikh > Research Fellow, > Department of Community Medicine > Faculty of Health, > University of Troms?, > Norway > > > > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -- ==================================== Teppei Yamamoto Assistant Professor Department of Political Science Massachusetts Institute of Technology http://web.mit.edu/teppei/www/ ==================================== _______________________________________________ Mediation-information mailing list Mediation-information at lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -------------- next part -------------- An HTML attachment was scrubbed... URL: From teppei at mit.edu Wed Aug 27 18:02:05 2014 From: teppei at mit.edu (Teppei Yamamoto) Date: Wed, 27 Aug 2014 12:02:05 -0400 Subject: [Mediation-information] Causally dependent mediators In-Reply-To: <8564BCD7D26E0D40872F1A132C8BBB250258B234@MATIAEXCH.matiaf.local> References: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> <53FDDCA7.8060203@mit.edu> <8564BCD7D26E0D40872F1A132C8BBB250258B234@MATIAEXCH.matiaf.local> Message-ID: <53FE00FD.6040603@mit.edu> Hi Angel, The second model does not give you estimates of those effects "net of deprivation" -- you will get the direct and indirect effects of manipulating education levels from one level to another on aging. Including deprivation as a covariate in the models allows you to correct the confounding between education and economic position and/or education and aging due to deprivation, but that does not allow you to interpret the estimated mediation/direct effects as those effects "net of" deprivation. If you are interested in how the effect of deprivation on aging is transmitted by those two interrelated mediators, you want to use multimed (as in your original message) instead of model 2. Best, Teppei (8/27/14, 10:35 AM), Angel Rodriguez wrote: > > Thank you for your answers, Teppei. > > So, would these two models make sense? > > 1) deprivation --> education level --> aging well (covariates: age, > gender, region) > (with R mediation package, to study the direct and indirect effects of > deprivation on aging well) > > 2) education level --> economic position --> aging well (covariates: > DEPRIVATION, age, gender, region) > (with R mediation package, to study the direct and indirect effecs on > education on aging well, net of deprivation) > > Best, > > Angel > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > de Teppei Yamamoto > Enviado el: mi? 27/08/2014 15:27 > Para: Mediation-information at r-forge.wu-wien.ac.at > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > 25, Issue 3 > > Hi Angel, > > The difference between the two estimates (multimed vs. mediation > analysis on education level as the "treatment") is that the former is > about the effect of manipulating infant deprivation on aging whereas the > latter is about the effect of manipulating education directly. That is, > they are estimating effects of different kinds of hypothetical > interventions and how they are transmitted through the variables that > are assumed to be causally descendant. > > So I think your choice boils down to what causal quantity you are trying > to estimate. > > Best, > Teppei > > > (8/27/14, 5:38 AM), Angel Rodriguez wrote: > > Thank you again, Mashhood. > > > > What I do not understand yet is why we need a package like multimed if > > we can decompose a model with two causally dependent mediators (infant > > deprivation --> education level --> economic position --> aging well) > > into two models (infant deprivation --> education level --> aging well & > > education level --> economic position --> aging well). > > > > Following your advise, I should include infant deprivation as a > > confounder in the model education level --> economic position --> aging > > well. But if deprivation conditions education strongly, I could end not > > finding any direct nor indirect effect of education on aging well. > > > > Best regards, > > > > Angel Rodriguez-Laso > > > > > > -----Mensaje original----- > > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > > de Mashhood Sheikh > > Enviado el: mar 26/08/2014 20:56 > > Para: mediation-information at lists.r-forge.r-project.org > > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > > 25,Issue 3 > > > > > > Hi, > > > > In the model (education level --> economic position --> aging well) you > > should include all the variables that potentially confound the > > association between education level and economic position, education > > level and ageing well, and economic position and ageing well. > > > > Best wishes, > > Mashhood > > --- > > Mashhood Ahmed Sheikh > > Research Fellow, > > Department of Community Medicine > > Faculty of Health, > > University of Troms?, > > Norway > > > > > > > > > > _______________________________________________ > > Mediation-information mailing list > > Mediation-information at lists.r-forge.r-project.org > > > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > -- > ==================================== > Teppei Yamamoto > Assistant Professor > Department of Political Science > Massachusetts Institute of Technology > http://web.mit.edu/teppei/www/ > ==================================== > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > > > > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -- ==================================== Teppei Yamamoto Assistant Professor Department of Political Science Massachusetts Institute of Technology http://web.mit.edu/teppei/www/ ==================================== From senor_massao at hotmail.com Wed Aug 27 23:33:40 2014 From: senor_massao at hotmail.com (Mashhood Sheikh) Date: Wed, 27 Aug 2014 22:33:40 +0100 Subject: [Mediation-information] Mediation-information Digest, Vol 25, Issue 5 In-Reply-To: References: Message-ID: Hi, What I do not understand yet is why we need a package like multimed if we can decompose a model with two causally dependent mediators (infant deprivation --> education level --> economic position --> aging well) into two models (infant deprivation --> education level --> aging well & education level --> economic position --> aging well). I can not comment on this. Bit beyond my expertise.? But if deprivation conditions education strongly, I could end not finding any direct nor indirect effect of education on aging well. Yes, that may happen. The estimates will be biased if you exclude potential confounders.? --- Mashhood Ahmed Sheikh Research Fellow,? Department of Community Medicine, Faculty of Health, University of Troms?, Norway ---------------------------------------- > From: mediation-information-request at lists.r-forge.r-project.org > Subject: Mediation-information Digest, Vol 25, Issue 5 > To: mediation-information at lists.r-forge.r-project.org > Date: Wed, 27 Aug 2014 12:00:24 +0200 > > Send Mediation-information mailing list submissions to > mediation-information at lists.r-forge.r-project.org > > To subscribe or unsubscribe via the World Wide Web, visit > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > or, via email, send a message with subject or body 'help' to > mediation-information-request at lists.r-forge.r-project.org > > You can reach the person managing the list at > mediation-information-owner at lists.r-forge.r-project.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Mediation-information digest..." > > > Today's Topics: > > 1. Re: Mediation-information Digest, Vol 25, Issue 3 > (Angel Rodriguez) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Wed, 27 Aug 2014 11:38:53 +0200 > From: "Angel Rodriguez" > To: > Subject: Re: [Mediation-information] Mediation-information Digest, Vol > 25, Issue 3 > Message-ID: > <8564BCD7D26E0D40872F1A132C8BBB250258B231 at MATIAEXCH.matiaf.local> > Content-Type: text/plain; charset="iso-8859-1" > > Thank you again, Mashhood. > > What I do not understand yet is why we need a package like multimed if we can decompose a model with two causally dependent mediators (infant deprivation --> education level --> economic position --> aging well) into two models (infant deprivation --> education level --> aging well & education level --> economic position --> aging well). > > Following your advise, I should include infant deprivation as a confounder in the model education level --> economic position --> aging well. But if deprivation conditions education strongly, I could end not finding any direct nor indirect effect of education on aging well. > > Best regards, > > Angel Rodriguez-Laso > > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Mashhood Sheikh > Enviado el: mar 26/08/2014 20:56 > Para: mediation-information at lists.r-forge.r-project.org > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol 25,Issue 3 > > > Hi, > > In the model (education level --> economic position --> aging well) you should include all the variables that potentially confound the association between education level and economic position, education level and ageing well, and economic position and ageing well.? > > Best wishes, > Mashhood > --- > Mashhood Ahmed Sheikh > Research Fellow, > Department of Community Medicine > Faculty of Health, > University of Troms?, > Norway > > > -------------- next part -------------- > An HTML attachment was scrubbed... > URL: > > ------------------------------ > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > End of Mediation-information Digest, Vol 25, Issue 5 > **************************************************** From angel.rodriguez at matiainstituto.net Thu Aug 28 09:44:40 2014 From: angel.rodriguez at matiainstituto.net (Angel Rodriguez) Date: Thu, 28 Aug 2014 09:44:40 +0200 Subject: [Mediation-information] Causally dependent mediators References: <8564BCD7D26E0D40872F1A132C8BBB250258B231@MATIAEXCH.matiaf.local> <53FDDCA7.8060203@mit.edu> <8564BCD7D26E0D40872F1A132C8BBB250258B234@MATIAEXCH.matiaf.local> <53FE00FD.6040603@mit.edu> Message-ID: <8564BCD7D26E0D40872F1A132C8BBB250258B235@MATIAEXCH.matiaf.local> Thank you Teppei and Mashhood for your very helpful comments. I would be using then: 1) deprivation --> education level --> aging well (covariates: age, gender, region) (with the R mediation package, to study the direct and indirect effects of deprivation on aging well) 2) deprivation --> education level --> economic position --> aging well (covariates: age, gender, region) (with the R MULTIMED package, to study the direct and indirect effecs mediated through economic position of deprivation on aging well) Does it seem OK? Best regards, Angel -----Mensaje original----- De: mediation-information-bounces at lists.r-forge.r-project.org en nombre de Teppei Yamamoto Enviado el: mi? 27/08/2014 18:02 Para: mediation-information at r-forge.wu-wien.ac.at Asunto: Re: [Mediation-information] Causally dependent mediators Hi Angel, The second model does not give you estimates of those effects "net of deprivation" -- you will get the direct and indirect effects of manipulating education levels from one level to another on aging. Including deprivation as a covariate in the models allows you to correct the confounding between education and economic position and/or education and aging due to deprivation, but that does not allow you to interpret the estimated mediation/direct effects as those effects "net of" deprivation. If you are interested in how the effect of deprivation on aging is transmitted by those two interrelated mediators, you want to use multimed (as in your original message) instead of model 2. Best, Teppei (8/27/14, 10:35 AM), Angel Rodriguez wrote: > > Thank you for your answers, Teppei. > > So, would these two models make sense? > > 1) deprivation --> education level --> aging well (covariates: age, > gender, region) > (with R mediation package, to study the direct and indirect effects of > deprivation on aging well) > > 2) education level --> economic position --> aging well (covariates: > DEPRIVATION, age, gender, region) > (with R mediation package, to study the direct and indirect effecs on > education on aging well, net of deprivation) > > Best, > > Angel > > -----Mensaje original----- > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > de Teppei Yamamoto > Enviado el: mi? 27/08/2014 15:27 > Para: Mediation-information at r-forge.wu-wien.ac.at > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > 25, Issue 3 > > Hi Angel, > > The difference between the two estimates (multimed vs. mediation > analysis on education level as the "treatment") is that the former is > about the effect of manipulating infant deprivation on aging whereas the > latter is about the effect of manipulating education directly. That is, > they are estimating effects of different kinds of hypothetical > interventions and how they are transmitted through the variables that > are assumed to be causally descendant. > > So I think your choice boils down to what causal quantity you are trying > to estimate. > > Best, > Teppei > > > (8/27/14, 5:38 AM), Angel Rodriguez wrote: > > Thank you again, Mashhood. > > > > What I do not understand yet is why we need a package like multimed if > > we can decompose a model with two causally dependent mediators (infant > > deprivation --> education level --> economic position --> aging well) > > into two models (infant deprivation --> education level --> aging well & > > education level --> economic position --> aging well). > > > > Following your advise, I should include infant deprivation as a > > confounder in the model education level --> economic position --> aging > > well. But if deprivation conditions education strongly, I could end not > > finding any direct nor indirect effect of education on aging well. > > > > Best regards, > > > > Angel Rodriguez-Laso > > > > > > -----Mensaje original----- > > De: mediation-information-bounces at lists.r-forge.r-project.org en nombre > > de Mashhood Sheikh > > Enviado el: mar 26/08/2014 20:56 > > Para: mediation-information at lists.r-forge.r-project.org > > Asunto: Re: [Mediation-information] Mediation-information Digest, Vol > > 25,Issue 3 > > > > > > Hi, > > > > In the model (education level --> economic position --> aging well) you > > should include all the variables that potentially confound the > > association between education level and economic position, education > > level and ageing well, and economic position and ageing well. > > > > Best wishes, > > Mashhood > > --- > > Mashhood Ahmed Sheikh > > Research Fellow, > > Department of Community Medicine > > Faculty of Health, > > University of Troms?, > > Norway > > > > > > > > > > _______________________________________________ > > Mediation-information mailing list > > Mediation-information at lists.r-forge.r-project.org > > > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > -- > ==================================== > Teppei Yamamoto > Assistant Professor > Department of Political Science > Massachusetts Institute of Technology > http://web.mit.edu/teppei/www/ > ==================================== > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information > > > > > > _______________________________________________ > Mediation-information mailing list > Mediation-information at lists.r-forge.r-project.org > https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -- ==================================== Teppei Yamamoto Assistant Professor Department of Political Science Massachusetts Institute of Technology http://web.mit.edu/teppei/www/ ==================================== _______________________________________________ Mediation-information mailing list Mediation-information at lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/mediation-information -------------- next part -------------- An HTML attachment was scrubbed... URL: