Help with multilevel cox regression

[Elvire Landstra]

Hello everyone,

I'm trying to evaluate the association between outcome CVD and exposure saturated fat intake. The data is multilevel (households were selected) so I used the -mestreg- command for the multilevel cox regression, specifying the outcome with -stset-. I used the following command:

mestreg exposure1 confounders || household:, distribution(weibull)

However, I have simply assumed that the data follows a weibull distribution (as it is survival data). I would like to test whether this is true. I've read somewhere that you can use the -wbull- command, but am unsure how to run and interpret it. Should I run it on only the households (so 1 person per household) or should I run it on all individuals?

I have tried running it on the individuals in the sample and got this result:

. wbull id

Fitting Weibull distribution to id

b 6305.46272
c 1.62174

(id is simply the unique person id, so every person in the dataset has one).

Is the -wbull- command the correct command to test the weibull distribution? Is there another way to test that the data fits the weibull distribution?

Thank you in advance,

Elvire Landstra

[Marcos Almeida]

If I understood right - wbull - is a user-written program whose author is Nick Cox.

By read the abstract of the program, we have this:

wbull fits a Weibull distribution to a single variable by the
method of maximum likelihood. Frequency weights may be used to
fit variables given in frequency distribution form. The
distribution is here parameterised by a scale parameter b and a
shape parameter c.

[Carlo Lazzaro]

Elvire:
as an aside to Marcos' helpful advice, please note that while Cox regression is semiparametric, Weibull is parametric. You can take a look at -stcox- and -streg- entries in Stata .pdf manual to get a comprehensive description of these survival analysis models.
Stata users who deal with regression model for survival analysis (and survival analysis at large) like the following, really valuable textbook: https://www.stata.com/bookstore/surv...-introduction/