Re: st: Calculating pooled estimates using random-effects logistic regression

[Alfonso Sánchez-Peñalver <alfonso.statalist@gmail.com>]

Hi,

I don't understand what you mean by pooled estimates, since from the data you show each observation is a different study, and no study is repeated in more than one observation. I also don't understand what kind of model you want to estimate here with a logistic regression, since you don't have an explanatory variable, as the number of patients ought to be what you refer to as sample size and thus used as weights.

Having said that, if you were to have more variables that you could use to estimate a logistic regression model, and since your response variable (survival rate) is a fractional response variable, you could consistently estimate the logistic model using the -glm- command with the -familyname- option set to binomial, and the -linkname- option set to logit, since -glm- accepts weights. For examples type -help glm-. The seminal paper on simple (not pooled) fractional response variables is:

Papke, Leslie E. and Jeffrey M. Wooldridge (1996), "Econometric Methods For Fractional Response Variables With An Application to 401(K) Plan Participation Rates," Journal of Applied Econometrics, Vol. 11, 619-632.

When you indeed have a pooled (where certain units of interest are observed more than once) dataset with a fractional response variable, the paper to read is

Papke, Leslie E. and Jeffrey M. Wooldridge (2008), "Panel data methods for fractional response variables with an application to test pass rates," Journal of Econometrics, Vol. 145, 121-133

Best,

Alfonso Sanchez-Penalver, PhD.

Visiting Assistant Professor
Suffolk University

Senior Instructor
University of Massachusetts, Boston


On Feb 17, 2014, at 9:21 AM, Philipp Do <q017mm@gmail.com> wrote:

> I want to calculate pooled estimates of 1-year survival rates from
> several studies using random-effects logistic regression analysis
> (sample weighting should be applied according to the sample size).
> What is the most appropriate way to perform this analysis with Stata?
> Can you give me an example on the appropriate command using the data
> below?
> 
> Study   No. of patients     1-year survival rate (%)
> #1                46                  40
> #2                21                  31
> #3                36                  47
> #4                25                  33
> #5                11                  27
> #6                16                  39
> #7                46                  37
> #8                29                  22
> #9                62                  30
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/