{smcl}
{hline}
help for {hi:wclogit}
{hline}
{title:Conditional logistic regression with within-group varying weights}
{p 8 27}
{cmdab:wclogit}
[{it:varlist}] [{cmd:using}] [{hi:if}{it: exp}] [{hi:in}{it: exp}] [{help:weight}{it: exp}]
,
{cmdab:g:roup}{cmd:(}{it:varname}{cmd:)}
[
{cmdab:l:evel}{cmd:(}{it:#}{cmd:)}
{cmd:or}
]
{p}
{title:Description}
{p 0 0}
This command maximises the partial log-likelihood for conditional logistic regression with
weights that can vary within the matched set defined by the {hi:group} option. In calculations
the Breslow likelihood as specified in the STATA manual. Note that only one case is
allowed within one group so no ties are possible and that frequency weights are not allowed.
In other words the dataset should be expanded to an "individual data" format.
{p 0 0}
The main application for this function would be when you have inverse probability weights
to correct for any bias introduced in selecting members of each group. A typical
epidemiological design would be the case-cohort study where every case from a larger cohort
is included in the analysis but a smaller random sample of all possible controls are included.
{title:Options}
{p 0 4}
{cmdab:g:roup}{cmd:(}{it:varname}{cmd:)} is not optional; it specifies the identifier variable (numeric or string) for the matched groups.
{p 0 4}
{cmdab:l:evel}{cmd:(}{it:#}{cmd:)} specifies the confidence level, in percent, for the confidence intervals of the coefficients; see help {help level}.
{p 0 4}
{cmd:or} reports the estimated coefficients transformed to odds ratios, i.e., exp(b)
rather than b. Standard errors and confidence intervals are
similarly transformed. This option affects how results are displayed, not how they
are estimated. {hi: or} may be specified at estimation or when
redisplaying previously estimated results.
{title:Examples}
{p 0 0}
To analyse an "enriched" matched case control study using inverse probability weights
specified in the variable {hi:ipw}.
{inp: .wclogit case exp [aw=ipw], or level(99) group(matchedset)}
{p 0 0}
Although in this example the option used "analytical weights" [aw=ipw] it is because these seemed
more natural for this command. Inverse probability weights are definitely not frequency or probability weights.
{title:Author}
{p}
Adrian Mander, MRC Human Nutrition Research Unit, Cambridge, UK.
Email {browse "mailto:adrian.mander@mrc-hnr.cam.ac.uk":adrian.mander@mrc-hnr.cam.ac.uk}
{title:Also see}
On-line: help for {help clogit}, {help stcox}.