Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: impute with draws from random distribution

From   Richard Goldstein <>
Subject   Re: st: impute with draws from random distribution
Date   Wed, 22 Jun 2011 08:24:26 -0400

how were participants chosen? e.g., if chosen randomly, then there must
have been a pool from which they were chosen and a time at which people
were chosen (though that time might have varied and the composition of
the pool might have varied by time); if not chosen randomly, could the
non-participants have been chosen? if no, you have a potential problem
(depending on why the answer is "no"); if yes, then you can still do
something w.r.t. who was in what pool and when

hope this helps,


On 6/22/11 8:13 AM, D-Ta wrote:
> i totally agree, it doesnt give any additional information. BUT, it
> would simply help me to readjust the outcomes compared between treatment
> and control group relative to the start-date of the programm.
> And to do this I need a hypothetical start-date for non-participants.
> How else could you estimate (e.g.) the treatment effect of training on
> (e.g.) the employment probability (e.g.) 2 month after training start?
> Thats what I am after.
> Thanks for your support
> Darjusch
> Am 22.06.2011 14:01, schrieb Maarten Buis:
>> On Wed, Jun 22, 2011 at 1:46 PM, D-Ta<>  wrote:
>>> So, whats a valid comparison group then in your opinion, taking into
>>> account
>>> the timing of the training? Or should I neglect that information
>>> about the
>>> timing completing when matching?
>> As you said yourself it is logically impossible to get someone with
>> the same timing of the intervention when one group did and another
>> group did not receive the intervention. So you cannot use that
>> variable to match.
>> Look at this problem differently: What would you be matching on when
>> you imputed timing? Imputation does not create new information, it
>> just replicates the patterns that are present in the observed part of
>> the data. So imputing timing and than matching on imputed timing does
>> not add anything compared to just leaving timing out and match on the
>> variables you would have used to impute.
>> Hope this helps,
>> Maarten
>> --------------------------
>> Maarten L. Buis
>> Institut fuer Soziologie
>> Universitaet Tuebingen
>> Wilhelmstrasse 36
>> 72074 Tuebingen
>> Germany
>> --------------------------
*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index