Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Controlling for atrittion in panel, or creating a panel without atrittion


From   Laurie Molina <molinalaurie@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Controlling for atrittion in panel, or creating a panel without atrittion
Date   Thu, 23 Feb 2012 09:35:56 -0600

Thank you brendan!

On Thu, Feb 23, 2012 at 4:31 AM, Brendan Halpin <brendan.halpin@ul.ie> wrote:
> On Thu, Feb 23 2012, Laurie Molina wrote:
>
>> 1. To take a random sample at the first period and then merge the
>> databases to have a panel. In this case my panel will have attrition.
>> In fact by the last period the attrition rate (comparing period 1
>> observations with period 10 observations will be of almost 50%). The
>> remaining sample size much smaller than the original sample size is
>> leading to representativeness issues.
>
> Yes, your sub-sample will have attrition bias, but in a way that
> reflects that of your main sample. That is, if your main sample is
> representative in wave 1 and then suffers from attrition, this is
> probably a good option, but if subjects can meaningfully enter the main
> sample after wave 1 perhaps not.
>
>
>> 2. To take a random sample of the observations that appear on the
>> database in all the periods. If attrition is not random, then i would
>> have a population that is different to the original population, and
>> hence my random sample may not be representative for the original
>> population, but only for the population defined by all the
>> observations that appear in the database in all periods.
>
> Depending on what analysis you do afterwards, you may end up imposing
> this restriction anyway.
>
>> Which option do you think i should take?
>
> A third option would be to take a random sample of IDs, and take all
> observations available for them. This will better replicate the pattern
> of missingness in the main sample.
>
> Which option is best depends on what analysis you want to do, and on how
> the main sample is collected (in particular, in what way and at what
> time it is representative),
>
> Brendan
> --
> Brendan Halpin,   Department of Sociology,   University of Limerick,   Ireland
> Tel: w +353-61-213147  f +353-61-202569  h +353-61-338562;  Room F1-009 x 3147
> mailto:brendan.halpin@ul.ie    ULSociology on Facebook: http://on.fb.me/fjIK9t
> http://teaching.sociology.ul.ie/bhalpin/wordpress         twitter:@ULSociology
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index