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: Fwd: help with xtmixed and margins


From   "Visintainer, Paul" <Paul.Visintainer@baystatehealth.org>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: Fwd: help with xtmixed and margins
Date   Tue, 25 Sep 2012 08:27:10 -0400

Jim,

Two possibilities:  first, the tests in margins are coming from an estimation command -- xtmixed.  Thus, any tests done as post-estimation will be accomplished within a specific covariance structure which includes other covariates in the model.  The t-test command doesn't account for the covariance structure.

Secondly, the margins command uses a "delta-method" for estimating the standard errors and uses a z-distribution.  Below are two posts that deal with this issue.

http://www.stata.com/statalist/archive/2010-12/msg00597.html

http://www.stata.com/statalist/archive/2010-12/msg00597.html


-Paul


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of James prince
Sent: Tuesday, September 25, 2012 6:37 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Fwd: help with xtmixed and margins

Paul,thank you for your suggestions. Regarding the t-test, I did
notice that MARGNS outputs SE. I converted them to SD (i.e.
sqrt(N)*se) and use the output in TTESTI. Notice in the output from
the t-test that the SE are the same as those reported by MARGIN,
however the p-values are very different from that reported by the TEST
command after MARGINS, POST. What am I missing?

Thank you again,
Jim

On Mon, Sep 24, 2012 at 1:44 PM, Visintainer, Paul
<Paul.Visintainer@baystatehealth.org> wrote:
> Jim,
>
> The first thing to note is that, in your t-test command, you are using standard errors rather than standard deviations.  Because the -margins- command is based on the prior estimation command, it outputs standard errors.
>
> Secondly, -margins- is a powerful, but complex command.  I strongly recommend spending a lot time with the help file and its examples. (My printed copy is well-worn.)  Also, -margins- has options to run the types of tests you want to do -- either with options or with related commands: e.g., -contrast- and -pwcompare-.
>
> I would also recommend getting some of the books on Stata programming (e.g., Sophie Rabe-Hesketh series on multilevel models and Michael Mitchell's book on visualizing regression models are very helpful.  There are several other great texts at the Stata bookstore).  These have a multitude of useful examples and commands.
>
> Paul
>
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of James prince
> Sent: Monday, September 24, 2012 1:10 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Fwd: help with xtmixed and margins
>
> I did not get a reply on my post from last week. So let me ask a
> simpler question.
>
> After a simple XTMIXED (intercept model and no interaction terms),
> what is the difference between:
>
> margins overweight, over( month)
>
> and
>
> margins, over( month overweight)
>
> They yield different results.
>
> Thank you,
>
> Jim
>
> On Fri, Sep 21, 2012 at 8:32 AM, James prince <jsprince26@gmail.com> wrote:
>> Hi,
>>
>> I am new to Stata XTMIXED command and having some trouble. I hope
>> someone could help.
>> I am using the command MARGINS to predict the fitted means for what I
>> thought was a simple mixed model (random intercept only).
>> I have longitudinal data on children collected a 3, 6, 9, 12 and 24 months.
>> I'm interested in the relationship between a physical activity score
>> (score) and weight (dichotomized as overweight, yes, no).
>> I'm treating "month" as discrete, and will like to test if the mean
>> fitted score differs between by overweight status within month.
>> Here is what I did:
>>
>> . noi xtmixed score i.overweight i.race i.month  bwt bln i.nedu
>> i.gender || id:,mle
>>
>> <...>
>>
>>   I then use MARGINS with POST:
>> . margins overweight,over( month) post
>>
>> Predictive margins                                Number of obs   =       1748
>>
>> Expression   : Linear prediction, fixed portion, predict()
>> over         : month
>>
>> ----------------------------------------------------------------------------------
>>                  |            Delta-method
>>                  |     Margin   Std. Err.      z    P>|z|     [95%
>> Conf. Interval]
>> -----------------+----------------------------------------------------------------
>> month#overweight |
>>             3 0  |   100.9859   .4955637   203.78   0.000     100.0146
>>    101.9572
>>             3 1  |   99.36709   .6820952   145.68   0.000     98.03021
>>    100.704
>>             6 0  |   103.8597   .5121105   202.81   0.000     102.8559
>>     104.8634
>>             6 1  |   102.2409   .6479719   157.79   0.000     100.9709
>>   103.5109
>>             9 0  |   97.75468   .5267714   185.57   0.000     96.72223
>>   98.78714
>>             9 1  |   96.13588    .636223   151.10   0.000     94.88891
>>   97.38285
>>            12 0  |   102.2707   .5738975   178.20   0.000     101.1458
>>   103.3955
>>            12 1  |   100.6519   .6700738   150.21   0.000     99.33853
>>   101.9652
>> ----------------------------------------------------------------------------------
>>
>> And then TEST to test difference at each month. For example at 3 months:
>> . test 3.month#0.overweight= 3.month#1.overweight
>>
>>  ( 1)  3bn.month#0bn.overweight - 3bn.month#1.overweight = 0
>>
>>            chi2(  1) =    7.05
>>          Prob > chi2 =    0.0079
>>
>> However, when I do a t-test using the values form margin, a very
>> different answer is observed:
>>
>> . ttesti 373 100.9859 9.5709248 96 99.36709 6.6831408
>>
>> Two-sample t test with equal variances
>> ------------------------------------------------------------------------------
>>          |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
>> ---------+--------------------------------------------------------------------
>>        x |     373    100.9859    .4955637    9.570925    100.0114    101.9604
>>        y |      96    99.36709    .6820952    6.683141    98.01296    100.7212
>> ---------+--------------------------------------------------------------------
>> combined |     469    100.6545    .4189191    9.072283    99.83135    101.4777
>> ---------+--------------------------------------------------------------------
>>     diff |             1.61881    1.036685                -.418334    3.655954
>> ------------------------------------------------------------------------------
>>     diff = mean(x) - mean(y)                                      t =   1.5615
>> Ho: diff = 0                                     degrees of freedom =      467
>>
>>     Ha: diff < 0                 Ha: diff != 0                 Ha: diff > 0
>>  Pr(T < t) = 0.9405         Pr(|T| > |t|) = 0.1191          Pr(T > t) = 0.0595
>>
>>
>> So, I have two questions:
>> 1. How can I compare the adjusted scores by overweight status for each month?
>> 2. What is the difference between these two MARGIN commands? They
>> yield different results:
>>
>> . margins overweight, over( month)
>> and
>> . margins, over( month overweight)
>>
>> Thank you,
>> James.
>>
>> *
>> *   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/
>
> ----------------------------------------------------------------------
> Please view our annual report at http://baystatehealth.org/annualreport
>
>
> CONFIDENTIALITY NOTICE: This e-mail communication and any attachments may contain confidential and privileged information for the use of the designated recipients named above. If you are not the intended recipient, you are hereby notified that you have received this communication in error and that any review, disclosure, dissemination, distribution or copying of it or its contents is prohibited. If you have received this communication in error, please reply to the sender immediately or by telephone at 413-794-0000 and destroy all copies of this communication and any attachments. For further information regarding Baystate Health's privacy policy, please visit our Internet site at http://baystatehealth.org.
>
> *
> *   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/

----------------------------------------------------------------------
Please view our annual report at http://baystatehealth.org/annualreport
 

CONFIDENTIALITY NOTICE: This e-mail communication and any attachments may contain confidential and privileged information for the use of the designated recipients named above. If you are not the intended recipient, you are hereby notified that you have received this communication in error and that any review, disclosure, dissemination, distribution or copying of it or its contents is prohibited. If you have received this communication in error, please reply to the sender immediately or by telephone at 413-794-0000 and destroy all copies of this communication and any attachments. For further information regarding Baystate Health's privacy policy, please visit our Internet site at http://baystatehealth.org.

*
*   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