I typically assess fit with a C-statistic; other
methods exist but can have counterintuitive properties.
Just be sure to round the predicted values into a
manageable number of classifications (I ususally 100).
xtmelogit pvar ...
predict phat
replace phat=round(phat,0.01)
roctab pvar phat
This reports the ROC area, also known as the C-statistic,
along with the 95%CI. A C-stat of 1 is a perfect fit model,
a C-stat of 0.5 is no better than chance. One advantage,
beyond simple interpretability (the C-stat assesses how
well the model predicts true outcomes), is that it is directly
comparable across very different model specifications.
hth,
Jeph
Hardy, Dale S wrote:
Does anyone know how to do model fit for xtmelogit models?
Dale
Dale Hardy, PhD RD CDE CHES
Research Associate
University of Texas School of Public Health
Houston, TX 77030
Phone: (713) 500-9221
Fax: (713) 500-9264
Email: [email protected]
*
* 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/
