I have written a program which generates a table of odds ratios from many
bivariate logistic regression models. I would like to add the p values
associated with the Hosmer Lemeshow chi2 statistic (as generated by
. estat gof, group(10)
to the table. -estat- returns r(chi2) but not the p-value. The stata
function 'chi2' returns a very different p-value than that returned by
-estat gof- Any ideas?
You should have used -chi2tail- instead of -chi2-.
*----------begin example-------
sysuse auto, clear
recode rep78 1=2
tab rep78 foreign, chi2
return list
di chi2(3,r(chi2))
di 1-chi2(3,r(chi2))
di chi2tail(3,r(chi2))
*-----------end example------------
The bivariate logistic models consist of one continuous dependent
variable
of interest which is included in every model (var1, below) together with
a
different (mostly dichotomous) variable (var2) for each model, i.e.:
.logistic death var1 var2
The point is to assess the change in the odds ratio associated with var1
with inclusion of var2.
It looks like you are doing some sensitivity analysis or extreme bounds
analysis like
(Sala-i-Martin 1997), is that right?
Depending on var2, .estat gof, group(10) will generate less than 10
quantiles due to ties.
The question: In reporting multiple p values for multiple gof tests,
should
I accept whatever groupings Stata gives me and report the stats with
different df's for each one? Or should I force the same # of groups for
each model?
I don't see why you would use the Hosmer Lemeshow chi2 statistic if it is
inconvenient. You are
after all interested in changes in the parameter of interest, not in the
goodness of fit. More
specifically, Agresti (2002, p. 177) warns that the Hosmer-Lemeshow
statistic does not have good
power for detecting particular types of lack of fit, i.e. you accept
models as having a good fit
(or more precisely you do not reject the hypothesis that the fit is bad),
which should not be
accepted. If you want to report a goodness of fit statistic I would report
the BIC, since this is
comparabel across non-nested models. See for instance (Raftery 1995) (and
discussants in the
following articles) for a discussion. As an added bonus Adrian Raftery
promotes the use of BICs
for approximate Bayesian Model Averaging, which is not unlike
Sala-i-Martin's (2003) later take on
this model uncertainty problem.
Sala-i-Martin, Xavier, 1997. "I Just Ran Two Million Regressions,"
American Economic Review,
American Economic Association, vol. 87(2), pages 178-83.
Sala-i-Martin, X, G. Doppelhoffer, and R. Miller, "Determinants of Long
Term Growth: A Bayesian
Average of Classical Estimates (BACE) Approach", American Economic Review,
September 2003.
Agresti, Alan, (2002) Categorical Data Analysis, 2nd edition, Wiley
Raftery, Adrian E. (1995). Bayesian model selection in social research
(with Discussion).
Sociological Methodology, 25, 111-196.
HTH,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z214
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
___________________________________________________________
Yahoo! Photos - NEW, now offering a quality print service from just 8p a
photo http://uk.photos.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/