Re: st: Re: simple way to create missing data that is "missing atrandom" from a small datset

[Suzy <scott_788@wowway.com>]

Thank you Maarten. What I also did is dichotomize bmi  missingness - 
(generated newvar bmicat = 1 missing ; 0 otherwise). I then ran a 
logistic regressions with bmicat as the binary response variable  
univariately (age alone, sex alone, race alone, etc...) and then with 
the full  model. In each case, the odds of BMI missingness was 
significantly associated with age, but not with any other variables. Age 
was even associated with bmicat in the full model after accounting for 
the other variables). I heard that this is an approach that can be used 
to assess MCAR vs. MAR. Do you agree?

tab bmicat

     bmimi |      Freq.     Percent        Cum.
------------+-----------------------------------
         0 |        305       91.87       91.87
         1 |         27        8.13      100.00
------------+-----------------------------------
     Total |        332      100.00


logistic bmicat age sex fhdm dmcat race

Logistic regression                               Number of obs   
=        332
                                                 LR chi2(5)      =      
37.96
                                                 Prob > chi2     =     
0.0000
Log likelihood = -74.639705                       Pseudo R2       =     
0.2028

------------------------------------------------------------------------------
      bmicat | Odds Ratio   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
        age |   1.121633   .0268219     4.80   0.000     1.070276    
1.175454
        sex |   .9201524   .4542155    -0.17   0.866     .3496878    
2.421247
       fhdm |   1.060376   .5558315     0.11   0.911     .3795549    
2.962413
      dmcat |   .7724482   .4741646    -0.42   0.674     .2319329    
2.572625
       race |   1.340202    .791231     0.50   0.620     .4213434    
4.262891
------------------------------------------------------------------------------



Maarten buis wrote:


> Suzy:
> You wanted to create missingness according the to a MAR process, in your case the probability of
> missingness in the variable bmi should depend on the variable age. So we created the probability
> of missingness for each observation. The youngest person in your dataset has a probablity of
> missingness of invlogit(-8 + .1*28) = .0054863 (type -di invlogit(-8 + .1*28)-) and the oldest
> person has a missingness of invlogit(-8 + .1*82) = .549834. If the probability of missingness was
> constant (or random and unrelated to any of the other variables) than the missingness mechanism
> would be missing completely at random MCAR.
>
> HTH,
> Maarten
>
> --- Suzy <scott_788@wowway.com> wrote:
>  
>
> > I'm not sure what the implications are of the std dev and the 
> > max values of p (.549).
> >    
> -----------------------------------------
> between 1/2/2006 and 31/3/2006 I will be
> visiting the UCLA, during this time the
> best way to reach me is by email
>
> Maarten L. Buis
> Department of Social Research Methodology
> Vrije Universiteit Amsterdam
> Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
>
> visiting adress:
> Buitenveldertselaan 3 (Metropolitan), room Z214
>
> +31 20 5986715
>
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
>
>
> 	
> 	
> 		
> ___________________________________________________________ 
> Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail http://uk.messenger.yahoo.com
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
>
>
>  

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/