Statalist


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

RE: st: Interpretation of regressionmodel of ln-transformed variable


From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   RE: st: Interpretation of regressionmodel of ln-transformed variable
Date   Fri, 7 Nov 2008 13:50:49 -0000

It means that a GLM with reciprocal link and gamma family is apparently
a better fit to your data than a GLM with reciprocal link and (default)
Gaussian family. 

But, as should be clear, just because a model does converge does not
mean it is any good, let alone the best for your data. Yet other
right-skewed families might be good too; other kinds of model might
better. Without your data we can only speculate. 

Nick 
[email protected] 

roland andersson

Yes it converged with family(gamma) and after I had removed some
covariates. Can someone tell med what an analysis with family(gamma)
link(power -1) means?

I think this gets too complicated and the difference in LOS is so
small that it has no importance. I will content with stating there was
no difference in median LOS and use Mann-Whitney test and skip the
modelling of the LOS.

2008/11/6 Nick Cox <[email protected]>:

> Sometimes a gamma family works well for reciprocal links.

 roland andersson

> Thank you Peter
>
> I tried that but unfortunately the model did not converge.

2008/11/6 Lachenbruch, Peter <[email protected]>:

>> You need a space between power and -1
>> Check the help for glm

Roland andersson

>> I tried the model
>>
>> xi: glm lengthof stay lapscopy i.appdgn2 i.alderk prepermalign
>> precardioscleros  preperdiabetes cons, eform link(power-1) nocons
>> and get an error message "unrecognized command:  power"

*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index