Monday, January 02, 2006

st: RE: cloglog loglog etc

Since I am at least partially responsible for authoring both the glm and cloglog commands (James Hardin and I wrote the current glm command and the cloglog is derived from my STB program) I feel it appropriate to respond. I have also reviewed SPSS several times as software reviews editor for The American Statistician (TAS). I have found many mistakes in SPSS in my past reviews. They are particurlarly poor in GLM-based models. Thay have no Poisson command (I commented on this several times in reviews) and have produced rather unorthodox output for logistic regression models. I'd have to look up my reviews, but I do recall more than several mistakes in this area of statistics. So it is no suprise to me for you to have discovered a proglem with cloglog.loglog. I double checked the glm, cloglog, and loglog output of Stata, and double checked the code. It is correct. If you don't mind, I'll check SPSS, confirm your findings, and mention your discovery in my next TAS Section Editor's Notes. I'll be pleased to advise readers that it was you who identifed the problem. Again, Stata's cloglog program is correct. Stata's glm command, with cloglog and loglog links, is also correct, with the cloglog link being is consistent (identical) to cloglog. If your SPSS results are as you say, SPSS is incorrect. I'll check it out and advise TAS readers -- as well as SPSS. Joe Hilbe

------------------------------

Date: Sat, 31 Dec 2005 22:09:44 -0500 From: Richard Williams <Richard.A.Williams.5@ND.edu> Subject: st: Terminology - cloglog, loglog, nloglog

I've noticed an inconsistency in terminology between SPSS and Stata. I assume that Stata is "right", but I was wondering if anybody can shed more light on this.

PLUM is SPSS's ordinal regression routine. It supports various links, which it calls logit, probit, cloglog (complementary log log), and nloglog (negative log log).

Using a dichotomous DV, I've run the same models in PLUM and in Stata using the logit, probit, cloglog, and glm routines. What I have found is that

(1) SPSS and Stata produce the same results for links logit and probit

(2) SPSS's cloglog does not produce the same results as Stata's cloglog. Rather, cloglog in SPSS produces the same results as the loglog link in Stata's glm.

(3) SPSS's nloglog link is equivalent to Stata's cloglog link.

Now, it wouldn't be so bad if the 2 programs used different names for the same thing. But, in this case, they also use the same term, cloglog, for different things. Further, just looking over the formulas presented by SPSS & Stata, it isn't obvious to me that one has a more logical claim to the name cloglog than the other does. Any insights on this would be appreciated.

- ------------------------------------------- Richard Williams, Notre Dame Dept of Sociology OFFICE: (574)631-6668, (574)631-6463 FAX: (574)288-4373 HOME: (574)289-5227 EMAIL: Richard.A.Williams.5@ND.Edu WWW (personal): http://www.nd.edu/~rwilliam WWW (department): http://www.nd.edu/~soc * * 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/


Tag:


Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?