[FLR-list] FLCohort
Laurie
lauriekell at googlemail.com
Tue Oct 18 14:34:28 CEST 2011
On 10/18/2011 01:34 PM, ernesto.jardim at jrc.ec.europa.eu wrote:
> On 10/18/2011 01:22 PM, Laurie wrote:
>> On 10/18/2011 01:10 PM, Ernesto Jardim wrote:
>>>
>>>
>>> On Tue, Oct 18, 2011 at 1:03 PM, Laurie <lauriekell at googlemail.com
>>> <mailto:lauriekell at googlemail.com>> wrote:
>>>
>>> On 10/18/2011 12:50 PM, ernesto.jardim at jrc.ec.europa.eu
>>> <mailto:ernesto.jardim at jrc.ec.europa.eu> wrote:
>>>> On 10/18/2011 12:02 PM, Laurie wrote:
>>>>> On 10/18/2011 11:41 AM, ernesto.jardim at jrc.ec.europa.eu
>>>>> <mailto:ernesto.jardim at jrc.ec.europa.eu> wrote:
>>>>>> On 10/18/2011 11:26 AM, Laurie wrote:
>>>>>>> I am working on tagging models, where tags are released for
>>>>>>> a number of years and then recaptured in subsequent years.
>>>>>>> This allows M & Z to be estimated along a cohort.
>>>>>>>
>>>>>>> The data are in the form of number of fish tagged and
>>>>>>> recovered each year of a cohort. This means that you can´t
>>>>>>> use FLCohort which has age & cohort as dims 1&2.
>>>>>>>
>>>>>>> However, if you use the 1st dim for cohort then you can
>>>>>>> model this with an FLQuant, i.e.
>>>>>>>
>>>>>>> library(plyr)
>>>>>>>
>>>>>>> setGeneric("I", function(object,...)
>>>>>>> standardGeneric("I"))
>>>>>>> setGeneric('O', function(object, ...)
>>>>>>> standardGeneric("O"))
>>>>>>>
>>>>>>> setMethod('I', signature(object='FLQuant'),
>>>>>>> function(object,...){
>>>>>>> dmns <-dimnames(object)
>>>>>>> dmns[[1]]
>>>>>>> <-ac((dims(object)$minyear-dims(object)$max):(dims(object)$maxyear-
>>>>>>> dims(object)$min))
>>>>>>> names(dmns)[1]<-"quant"
>>>>>>> flc <-FLQuant(NA,dimnames=dmns)
>>>>>>>
>>>>>>> t. <-as.data.frame(object)
>>>>>>> t.$cohort <-t.$year-t.$age
>>>>>>> flc[]
>>>>>>> <-daply(t.,c("cohort","year","unit","season","area","iter"),function(x)
>>>>>>> sum(x$data))
>>>>>>>
>>>>>>> return(flc)})
>>>>>>>
>>>>>>> setMethod('O', signature(object='FLQuant'),
>>>>>>> function(object,...){
>>>>>>> dmns <-dimnames(object)
>>>>>>> dmns[[1]]
>>>>>>> <-ac((dims(object)$maxyear-dims(object)$max):(dims(object)$minyear-dims(object)$min))
>>>>>>> names(dmns)[1]<-"age"
>>>>>>> flc <-FLQuant(NA,dimnames=dmns)
>>>>>>>
>>>>>>> t. <-as.data.frame(object)
>>>>>>> t.$age <-t.$year-t.$quant
>>>>>>> t. <-t.[!is.na <http://is.na>(t.$data),]
>>>>>>> flc[]
>>>>>>> <-daply(t.,c("age","year","unit","season","area","iter"),function(x)
>>>>>>> sum(x$data))
>>>>>>>
>>>>>>> return(flc)})
>>>>>>>
>>>>>>> data(ple4)
>>>>>>>
>>>>>>> m1=I(m(ple4))
>>>>>>> m2=O(m1)
>>>>>>>
>>>>>>> It is probably best to derive a new class for this object
>>>>>>> just to break VPAs etc. But what to call it? Also how does
>>>>>>> it relate to FLCohort.
>>>>>>> Laurie
>>>>>>
>>>>>> You may change the age in FLCohort
>>>>>>
>>>>>> library(FLCore)
>>>>>> data(ple4)
>>>>>> flc <- FLCohort(catch.n(ple4))
>>>>>> quant(flc) <- "year"
>>>>>>
>>>>>> why not using year in the first dimension ? it should work
>>>>>> out of the box.
>>>>>>
>>>>>> EJ
>>>>>>
>>>>>>
>>>>> But you don´t get what you want.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> I(m(ple4))[1:10,1:10]
>>>>> An object of class "FLQuant"
>>>>> , , unit = unique, season = all, area = unique
>>>>>
>>>>> year
>>>>> quant 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
>>>>> 1947 0.1 NA NA NA NA NA NA NA NA NA
>>>>> 1948 0.1 0.1 <tel:1948%200.1%20%200.1> NA NA NA NA NA NA NA NA
>>>>> 1949 0.1 0.1 <tel:1949%200.1%20%200.1> 0.1 NA NA NA NA NA NA NA
>>>>> 1950 0.1 0.1 0.1 0.1 NA NA NA NA NA NA
>>>>> 19510.1 0.1 0.1 0.1 0.1 <tel:0.1%20%200.1%20%200.1%20%200.1%20%200.1> NA NA NA NA NA
>>>>> 1952 0.1 0.1 0.1 0.1 0.1 0.1 NA NA NA NA
>>>>> 1953 0.1 0.1 0.1 0.1 0.1 0.1 0.1 NA NA NA
>>>>> 1954 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 NA NA
>>>>> 1955 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 NA
>>>>> 1956 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
>>>>>
>>>>> units: NA
>>>>> > FLCohort(m(ple4))[1:10,1:10]
>>>>> An object of class "FLCohort"
>>>>> , , unit = unique, season = all, area = unique
>>>>>
>>>>> cohort
>>>>> age 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
>>>>> 1 NA NA NA NA NA NA NA NA NA 0.1
>>>>> 2 NA NA NA NA NA NA NA NA 0.1 0.1
>>>>> 3 NA NA NA NA NA NA NA 0.1 0.1 0.1
>>>>> 4 NA NA NA NA NA NA 0.1 0.1 0.1 0.1
>>>>> 5 NA NA NA NA NA0.1 0.1 0.1 0.1 0.1 <tel:0.1%20%200.1%20%200.1%20%200.1%20%200.1>
>>>>> 6 NA NA NA NA 0.10.1 0.1 0.1 0.1 0.1 <tel:0.1%20%200.1%20%200.1%20%200.1%20%200.1>
>>>>> 7 NA NA NA 0.1 0.1 0.1 0.1 0.1 0.1 0.1
>>>>> 8 NA NA 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
>>>>> 9 NA 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
>>>>>
>>>>> laurie
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>> Lost me. What do you want ?
>>>>
>>>> EJ
>>>>
>>>>
>>>>
>>> There are three ways to represent the 1st two dims in an FLQuant.
>>>
>>> i) age & year
>>> ii) age & cohort
>>> iii) cohort & year
>>>
>>> (i) is FLQuant and (ii) is FLCohort; However we do not have
>>> (iii) which is what I need for my tagging model.
>>>
>>> I have created an "FLCohortYear" from an FLQuant by adding I() &
>>> O() as coercion methods and validity that only allows 1st dim to
>>> be called cohort.
>>> But then we have two FLCohort objects which might conflict.
>>>
>>> Laurie
>>>
>>>
>>>
>>> But you can use FLQuant and change the first dimension. You may need
>>> a method to convert i) or ii) into it but shouldn't create any
>>> conflicts. If you really need a new data structure and the problem
>>> only affects tag data, call it FLTag and extend FLQuant.
>>>
>>> Best
>>> EJ
>>>
>> I would have called in FLCohort but some body got there 1st!
>>
>> Having 2 versions of FLCohorts with different properties does raise
>> an issue though.
>>
>> Laurie
>
> Yes, sure ! But you get what you need computing year=cohort+age, so I
> don't see the need for a new one.
>
> Best
>
> EJ
Same argument applies to FLQuant! However, the maths is easier in my
case using sweep & plyr if dims are year & cohort
Laurie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/flr-list/attachments/20111018/fe3d5dd1/attachment-0001.htm>
More information about the flr-list
mailing list