<div dir="ltr">Thanks. I tried running ictregBayes as well, but I've run into problems with the multiple sensitive item design.<div><br></div><div>Here's the code for what I've been trying to run. Note this is another multiple sensitive item list experiment that I can run ML on. It passes the design test too.</div>
<div>
<p class=""><br></p><p class="">ml.results<span class=""> </span><span class=""><-</span><span class=""> </span><span class="">ictreg(</span>full<span class=""> </span><span class="">~</span><span class=""> </span>male<span class=""> </span><span class="">+</span><span class=""> </span>lage<span class=""> </span><span class="">+</span><span class=""> </span>ordeduc<span class=""> </span><span class="">+</span><span class=""> </span>logincome<span class=""> </span><span class="">+</span>townsize<span class="">+</span>firmsize<span class="">,</span><span class=""> </span>data<span class=""> </span><span class="">=</span><span class=""> </span>true_subset<span class="">,</span><span class=""> </span>treat<span class=""> </span><span class="">=</span><span class=""> </span><span class="">"treatstat"</span><span class="">,</span><span class=""> </span>J<span class="">=</span><span class="">4</span><span class="">,</span><span class=""> </span>method<span class=""> </span><span class="">=</span><span class=""> </span><span class="">"ml", constrained=TRUE</span><span class="">)</span></p>
</div><div>
<p class="">draws<span class=""> </span><span class=""><-</span><span class=""> </span><span class="">mvrnorm(</span>n<span class="">=</span><span class="">3</span><span class="">,</span><span class=""> </span>mu<span class=""> </span><span class="">=</span><span class=""> </span><span class="">coef(ml</span>.results<span class="">),</span><span class=""> </span>Sigma<span class="">=vcov(ml</span>.results<span class="">)*</span><span class="">9</span><span class="">)</span></p>
</div><div>
<p class="">b1<span class=""> <- ictregBayes(</span>full<span class=""> </span><span class="">~</span><span class=""> </span>male<span class=""> </span><span class="">+</span><span class=""> </span>lage<span class=""> </span><span class="">+</span><span class=""> </span>ordeduc<span class=""> </span><span class="">+</span><span class=""> </span>logincome<span class=""> </span><span class="">+</span>townsize<span class="">+</span>firmsize<span class="">, </span>data<span class="">= </span>true_subset<span class="">, </span>treat<span class="">=</span><span class="">"treatstat"</span><span class="">, </span>J<span class="">=</span><span class="">4</span><span class="">,</span><span class=""> </span>n.draws<span class="">=100000</span><span class="">, </span>burnin<span class="">=</span>10000<span class="">, </span>thin<span class="">=</span><span class="">5</span><span class="">, </span>delta.start<span class="">=list(</span>draws<span class="">[1</span><span class="">,</span><span class="">1</span><span class="">:</span><span class="">7</span><span class="">],</span>draws2<span class="">[</span><span class="">3</span><span class="">,</span><span class="">8</span><span class="">:</span><span class="">14</span><span class="">]), </span>psi.start<span class="">=</span>draws1<span class="">[1</span><span class="">,</span><span class=""> 15</span><span class="">:</span><span class="">21</span><span class="">], </span>delta.tune<span class="">=diag(</span><span class="">0.00075</span><span class="">,</span><span class="">7</span><span class="">), </span>psi.tune<span class="">=diag(</span><span class="">0.00007</span><span class="">,</span><span class="">7</span><span class="">), </span>verbose<span class="">=</span><span class="">T, </span>constrained.multi = TRUE)</p>
<p class=""><span class=""><br></span></p><p class="">I keep getting the error: "Error in as.integer(ceiling) : 'ceiling' is missing"</p><p class=""><br></p><p class="">I also ran the sample code you provided in the help file using the 'race' data. The same error pops up when I run the multiple sensitive item design. I've also tried setting 'delta.tune' as a list with no luck either, as well as no luck with turning on and off the constrained function. We also looked into the package code and found the function ictregBayesMulti.fit which requires a ceiling parameter, but we couldn't figure out how to feed it one.</p>
<p class=""><br></p><p class="">Thanks for your help!</p><p class=""><br></p><p class="">David</p></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, Mar 3, 2014 at 8:24 PM, Kosuke Imai <span dir="ltr"><<a href="mailto:kimai@princeton.edu" target="_blank">kimai@princeton.edu</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Good. Our standard error calculation relies upon the numerical evaluation of hessian matrix, which in some cases is not very stable. We probably need to improve this in the next version but for now I think you can do the Bayesian model, which should be more stable. </div>
<div class="gmail_extra"><div class=""><br clear="all"><div><div dir="ltr"><div>Kosuke Imai</div><div>Department of Politics<br></div><div>Princeton University</div><div><a href="http://imai.princeton.edu" target="_blank">http://imai.princeton.edu</a></div>
</div></div>
<br><br></div><div><div class="h5"><div class="gmail_quote">On Mon, Mar 3, 2014 at 2:20 PM, David Szakonyi <span dir="ltr"><<a href="mailto:ds2875@columbia.edu" target="_blank">ds2875@columbia.edu</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div dir="ltr">Yes, it does pass the design test (p-value of 0.9384607). I'm going to try the Bayes model. Any other thoughts about why this would be happening?<div><br></div><div>Thanks,</div><div><br></div><div>David</div>
</div><div><div><div class="gmail_extra"><br><br><div class="gmail_quote">On Wed, Feb 26, 2014 at 9:41 AM, Kosuke Imai <span dir="ltr"><<a href="mailto:kimai@princeton.edu" target="_blank">kimai@princeton.edu</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Does your data pass the design test we've developed? Sometimes, the skewed data leads to this type of problems. Another possibility is to use a Bayesian model, which is available through ictregBayes().</div>
<div class="gmail_extra"><br clear="all"><div><div dir="ltr"><div>Kosuke Imai</div><div>Department of Politics<br></div><div>Princeton University</div><div><a href="http://imai.princeton.edu" target="_blank">http://imai.princeton.edu</a></div>
</div></div>
<br><br><div class="gmail_quote"><div><div>On Tue, Feb 25, 2014 at 6:44 PM, David Szakonyi <span dir="ltr"><<a href="mailto:ds2875@columbia.edu" target="_blank">ds2875@columbia.edu</a>></span> wrote:<br>
</div></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div><div>
<div dir="ltr">Hello!<div><br></div><div>I've been analyzing a two treatment list experiment that has been properly randomly assigned. </div><div><br></div><div>I just ran the following regression with one simple explanatory bivariate variable (gender), but received the following error:</div>
<div><font face="courier new, monospace"><br></font></div><div><div><font face="courier new, monospace">> ml.resultsnp <- ictreg(outcome ~ male, data = true_subsetnp, treat = "treatstatnp", J=4, method = "ml")</font></div>
<div><font face="courier new, monospace"><br></font></div><div><font face="courier new, monospace">Error in solve.default(-MLEfit$hessian) : </font></div><div><font face="courier new, monospace"> system is computationally singular: reciprocal condition number = 7.02376e-18</font></div>
</div><div><br></div><div>I then added an additional covariate (logged age), and the model converges, but with exploding standard errors:</div><div><br></div><div><div><font face="courier new, monospace">Item Count Technique Regression </font></div>
<div><font face="courier new, monospace"><br></font></div><div><font face="courier new, monospace">Call: ictreg(formula = outcome ~ male + lage, data = true_subsetnp, </font></div><div><font face="courier new, monospace"> treat = "treatstatnp", J = 4, method = "ml")</font></div>
<div><font face="courier new, monospace"><br></font></div><div><font face="courier new, monospace">Sensitive item (1)</font></div><div><font face="courier new, monospace"> Est. S.E.</font></div><div><font face="courier new, monospace">(Intercept) 3.94811 4.15077</font></div>
<div><font face="courier new, monospace">male -1.43966 1.21671</font></div><div><font face="courier new, monospace">lage -1.78966 1.14337</font></div><div><font face="courier new, monospace"><br></font></div>
<div><font face="courier new, monospace">Sensitive item (2)</font></div><div><font face="courier new, monospace"> Est. S.E.</font></div><div><font face="courier new, monospace">(Intercept) -7.93707 10.31522</font></div>
<div><b><font face="courier new, monospace">male -16.06552 2965.82521</font></b></div><div><font face="courier new, monospace">lage 0.99201 2.63957</font></div><div><font face="courier new, monospace"><br>
</font></div><div><font face="courier new, monospace">Control items</font></div><div><font face="courier new, monospace"> Est. S.E.</font></div><div><font face="courier new, monospace">(Intercept) 0.21267 0.25312</font></div>
<div><font face="courier new, monospace">male 0.00017 0.05277</font></div><div><font face="courier new, monospace">lage -0.04225 0.06603</font></div><div><font face="courier new, monospace"><br></font></div>
<div><font face="courier new, monospace">Log-likelihood: -2039.366</font></div><div><font face="courier new, monospace"><br></font></div><div><font face="courier new, monospace">Number of control items J set to 4. Treatment groups were indicated by '1' and '2' and the control group by '0'.</font></div>
<div><br></div><div>A variety of other model specifications return very similar results: either failing to converge/compute or returning point estimates and very large standard errors (sometimes nearly exactly the same values for not at all correlated variables).</div>
<div><br></div><div>Does anyone have any suggestions about what might be going wrong? </div><div><br></div><div>Thanks,</div><div><br></div><div>David</div><span><font color="#888888"><div><br></div><div><br>
</div>-- <br><div dir="ltr">David Szakonyi<br>
Ph.D Candidate - Comparative Politics<br>Columbia University<br><a href="mailto:ds2875@columbia.edu" target="_blank">ds2875@columbia.edu</a><br></div></font></span></div></div>
<br></div></div>_______________________________________________<br>
listpackage-discuss mailing list<br>
<a href="mailto:listpackage-discuss@lists.r-forge.r-project.org" target="_blank">listpackage-discuss@lists.r-forge.r-project.org</a><br>
<a href="https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/listpackage-discuss" target="_blank">https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/listpackage-discuss</a><br></blockquote></div><br></div>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div dir="ltr">David Szakonyi<br>Ph.D Candidate - Comparative Politics<br>Columbia University<br><a href="mailto:ds2875@columbia.edu" target="_blank">ds2875@columbia.edu</a><br>
</div>
</div>
</div></div></blockquote></div><br></div></div></div>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div dir="ltr">David Szakonyi<br>Ph.D Candidate - Comparative Politics<br>Columbia University<br><a href="mailto:ds2875@columbia.edu" target="_blank">ds2875@columbia.edu</a><br>
</div>
</div>