[R] Joint test

[Hein Goemans]

Dear John,

Thank you very much. Yes, I think this should do it. The basic hypothesis 
is to test whether the outcome of war (win, lose, draw) has a different 
effect on different regime types. In other words, are leaders of some 
regime types more "sensitive" to the outcome of war than others.

It seems to work great for one set of models, but I get an error message in 
another. I will try to figure this out. Thanks very much indeed for your 
help.

Best, Hein.


--On Friday, February 06, 2009 4:41 PM -0500 John Fox <jfox at mcmaster.ca> 
wrote:

> Dear Hein,
>
> I'm not entirely sure that I understand what you want to do, but I think
> that you want to test that the coefficient of milwinwar equals that of
> civwinwar, and similarly and simultaneously for two other pairs of
> coefficients. If so, you should be able to use the linear.hypothesis()
> function in the car package:
>
> linear.hypothesis(modelPG2, c("milwinwar = civwinwar", "millosewar =
> civlosewar", "mildrawwar = civdrawwar"))
>
> You should be able to get the same test from the wald.test() function that
> you tried, but I believe that you would have to build the hypothesis
> matrix manually.
>
> I hope this helps,
>  John
>
> ------------------------------
> John Fox, Professor
> Department of Sociology
> McMaster University
> Hamilton, Ontario, Canada
> web: socserv.mcmaster.ca/jfox
>
>
>> -----Original Message-----
>> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On
>> Behalf Of Hein Goemans
>> Sent: February-06-09 3:30 PM
>> To: r-help at stat.math.ethz.ch
>> Subject: [R] Joint test
>>
>> Dear All,
>>
>> I am estimating a Cox proportional hazard model, with several
>> interactions of the type a*z + a*y + a*x + b*z + b*y + b*x.
>>
>> I need to know if the first three (the "a"s) are jointly significantly
>> different from the last three (the "b"s). I have tried several
>> approaches, but have been unsuccessful.
>>
>> Here's the model, and the code I came up with, with the obvious
>> shortcomings.
>>
>>
>> modelPG2 <- coxph(Surv(t0, t, d)  ~ civilian + monarch + txmonarch
>>                                + civwar  + lngdpcap + growth
>>                                + tropen4 +  dopen4
>>                                + lnpop
>>                                + age0 + entry1 +  powtimes
>>                                + initiator2 + defender2 + inherit
>>                                + milwinsh + millosesh + mildrawsh +
>> milwinwar + millosewar + mildrawwar
>>                                + civwinsh + civlosesh + civdrawsh +
>> civwinwar + civlosewar + civdrawwar
>>                                + monwinsh + monlosesh + mondrawsh +
>> monwinwar + monlosewar + mondrawwar
>>                                + frailty(ccode), na.action=na.exclude,
>> data=LeaderPG.data,
>>
>> control=coxph.control(eps=1e-09,iter.max=100,outer.max=100))
>>
>> library(aod)
>>
>> # To test if Military Leaders are equally sensitive to the outcome of WAR
>> as Civilian leaders we need a JOINT test.
>> wald.test(b=coef(modelPG2), Sigma = vcov(modelPG2), Terms= c(19:21),
>> H0=c(-2.9101, 2.4028, -1.6504))
>>
>> # wald.test(b=coef(modelPG2), Sigma = vcov(modelPG2), Terms= c(19:21),
>> H0=c(0, 2.4028, 0))
>>
>> wald.test(b=coef(modelPG2), Sigma = vcov(modelPG2), Terms= c(25:27),
>> H0=c(-8.2330,2.3041,-0.2626))
>>
>> Any help would be very much appreciated.
>>
>> Hein Goemans.
>>
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>
>
>
>