.-
help for ^ccweight^
.-
Create a weighting variable for analysing a case-control study
--------------------------------------------------------------
^ccweight^ varlist
[^,s^tatus^(^varname^)^ ^p^weight^(^varname^)^
^c^pc^(^exp^)^ ]
Description
-----------
^ccweight^ takes, as input, a varlist whose distinct values correspond
to case groups, and a status variable (1 for cases, 0 for controls) in
the option ^status^. It creates, as output, a new variable, suitable
for use as a ^pweight^ variable when the case-control study is analysed
by regression with robust variances.
Options for use with ^powercv^
----------------------------
^status(^varname^)^ gives the name of a variable identifying cases.
Nonzero nonmissing values are interpreted as cases, zero values
as controls, and missing values as units of indeterminate
status.
^pweight(^varname^)^ gives the name of a variable to contain the output
weights. It is replaced if it exists, and generated otherwise.
It is set to zero for orphan cases and controls and units of
indeterminate status, to one for non-orphan cases, and, for
non-orphan controls, to the case/control ratio amongst units
in the same case group, multiplied by ^cpc^ (see below). If
absent, this option defaults to the name ^_pweight^.
^cpc(^cpc^)^ is a scalar, real or integer expression, giving the number
of controls per case to which the weights are adjusted. If
absent, it defaults to one. This option does not affect the
odds ratios or mean differences created by the regression, but
may improve the aesthetics if the ^pweight^ is tabulated.
Remarks
-------
This program is intended for use with ^logistic^ or other regression
programs in case-control studies. The estimates are calculated as if
there was a constant number of controls per case (given by ^cpc^) in
all strata of the matching criterion defined by the varlist. Odds ratios
(or other regression coefficients) are therefore calculated as for a
hypothetical population, in which the matching criterion is distributed
as for the (non-orphan) cases. The method is an alternative to
conditional logistic regression for analysing unmatched studies, or
matched studies with varying numbers of controls per case.
Examples
--------
. ^ccweight age sex,st(case) pweight(pw)^
. ^ccweight casegp,st(status) cpc(4)^
Author
------
Roger Newson, Imperial College School of Medicine, London, UK.
Email: ^r.newson@@ic.ac.uk^
Also see
--------
Manual: ^[R] logistic^, ^[R] clogit^.