stset id() and/or stcox cluster()

[Roberto Bakker]

Hello,

I have basic question for which I cannot find the answer - only indirectly*;
(The stcox model: with event death, repeated measures of covariates ; so repeated times nested within person (id))
when I include

id(id)

in

stset

, I do not need to include

cluster(id)

in

stcox

anymore?
and
when I do not include

id(id)

in

stset

, I do need to include

cluster(id)

in

stcox

?
Is my reasoning correct?
*in textes I read: id(id) without cluster(id), or opposite; not both id() and cluster()

Best regards,

Roberto

[Marcos Almeida]

You may wish to take a look at the exemples here in the Stata Manaul

[Roberto Bakker]

Thank you for your quick reply. The examples in the stata manual are neither clear - i.e. the text does not make clear cut when what combine and why; I have to read between the lines of the examples with some indirect info, e.g. the use of

vce(robust)

in

stcox

is being combined with

id()

in

stset

.
I tried this, and SEM are little different as expected. By the way, the interesting part is that

estat phtest

changes clearly, i.e. some non-sign become very significant. I think I will use these result as they are more conservative.
Best regards, Roberto

[Marcos Almeida]

The examples in the stata manual are neither clear - i.e. the text does not make clear cut when what combine and why; I have to read between the lines of the examples with some indirect info

I'm afraid a disagree. And I'm sure this is not only my point of view.

That said, I wonder whether you would find a text that, taking the term you use, "combines" so different concepts, such as SEM and Cox regression. What is more, the commands you showed are sparsed, sort of "little chunks", making it difficult to guess which models are being referred.

For example, - stset - is something to be discussed before delving into - stcox - and the checking of the PH assumption is not supposed to be discussed after - stset - but after - stcox - command.

Anyway, the FAQ clearly recommends to inform "exactly" what was typed. In #1 as well as in #3, hardly there is a line with a complete command to be found, let alone the absence of data set (full, abridged or mock) and output overall.

Please act according to the FAQ, This is the really the best way to get helpful replies.

[Roberto Bakker]

My apologies. I will look the FAQ better.

[Roberto Bakker]

I agree with "For example, - stset - is something to be discussed before delving into - stcox - and the checking of the PH assumption is not supposed to be discussed after - stset - but after - stcox - command" and was unclear in my text which caused confusion.
"