[R-sig-ME] missing data + explanatory variables

[Luca Borger]

Hello,

> I don't receive anything from E. Charpentier?!

see:
http://markmail.org/search/?q=missing%20data%20%2B%20explanatory%20variables%20:%20important%20complement#query:missing%20data%20%2B%20explanatory%20variables%20%3A%20important%20complement+page:1+mid:oykvzrctut3dtpbj+state:results

> The problem is that my response is always of this pattern:
> 0 0 0 1
> 0 0 0 0
> 0 1 NA NA
> That's wher I had missing data, after the default, I don't have any data 
> for
> that individual. And the variable is necessary a sequence of 0 with a
> possible 1 at the end.
>
> I'm trying to use something better than the logistic GLM at each time
> period.

How about survival analysis with censoring?


HTH


Cheers,

Luca




----- Original Message ----- 
From: "christophe dutang" <dutangc at gmail.com>
To: "David Duffy" <David.Duffy at qimr.edu.au>
Cc: <r-sig-mixed-models at r-project.org>
Sent: Friday, March 26, 2010 1:15 PM
Subject: Re: [R-sig-ME] missing data + explanatory variables


> Hello,
>
> 2010/3/26 David Duffy <David.Duffy at qimr.edu.au>
>
>> On Thu, 25 Mar 2010, Christophe Dutang wrote:
>>
>>  I try to model a binary response variable over a small period of time. 
>> The
>>>>> problem is that for some lines, the response is missing. In this 
>>>>> mailing
>>>>> list archive, I do not find response to this question.
>>>>>
>>>>>  fm1 <- lmer(Reaction ~ Days + (Days|Subject), sleepstudy)
>>> sleepstudy2 <- sleepstudy
>>> sleepstudy2[180, "Reaction"] <- NA
>>> fm2 <- lmer(Reaction ~ Days + (Days|Subject), sleepstudy2)
>>>
>>> What am I doing wrong?
>>>
>>
>> This worked perfectly, did it not?
>
> the observation number is reduced by 1.
>
>
>>  As Emmanuel Charpentier summarized, you can't use information from the
>> "lines" where the response is missing (expect for prediction, the BLUP)
>> unless you move to some type of multivariate model that explicitly models
>> the interrelationships of the covariates, like the multiple imputation
>> models or SEM.
>
> I don't receive anything from E. Charpentier?!
>
>
>> But in that case, all that happens is you gain a little bit of 
>> information
>> about the distributions of the covariates, which may or may not influence
>> your model for the response variable.
>>
>
> By "some lines are missing", I mean I don't have both the response and the
> covariate. Let me explain what I'm doing. I work on default of individual.
> Every 6 month for each indvidual the response variable equals to 1 if he
> defaults on the period, 0 otherwise. For each time period I also observe
> explanatory variables (age, job, marital status,...).
>
> The problem is that my response is always of this pattern:
> 0 0 0 1
> 0 0 0 0
> 0 1 NA NA
> That's wher I had missing data, after the default, I don't have any data 
> for
> that individual. And the variable is necessary a sequence of 0 with a
> possible 1 at the end.
>
> I'm trying to use something better than the logistic GLM at each time
> period.
>
> Thanks in advance for any advice
>
> Christophe
>
>
>>
>> PS what are the "lines" you refer to?  What is the actual problem you are
>> working on?
>>
>>
>> --
>> | David Duffy (MBBS PhD)                                         ,-_|\
>> | email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
>> | Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
>> | 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v
>>
>
>
>
> -- 
> Christophe DUTANG
> Ph. D. student at ISFA
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>