.-
help for ^epiconf^                         (STB-49: sbe27)
.-

Assessing confounding effects in epidemiological studies 
---------------------------------------------------------------

    ^epiconf^ yvar xvar [^if^ exp] [^in^ range],  [^con(^covlist^)^ ^cat(^covlist^)^
                ^m^odel^(logit^|^poisson^|^cox)^ ^nog^raph ^b^ackward
                ^e^xpos^(^varname^)^ ^d^ead^(^varname^)^ ^det^ail 
		^l^evel^(^#^)^ ^co^ef graph_options]
        
^epiconf^ calculates the change in effect estimates such as odds ratio, 
rate ratio, or hazard ratio after adjusting for a potential confounder 
using either "forward selection" or "backward deletion" strategies. 

^epiconf^ also calculates and graphs adjusted effect measures and 
their confidence intervals. The order of variables being selected is 
based on the "importance" (magnitude) of "change in estimate". 
  

Variables required
------------------

    ^yvar^ -- a binary outcome variable for logistic or poisson 
        regression; a survival time variable for Cox proportional hazards
        model.  

    ^xvar^ -- a binary exposure variable

         
Options 
-------
    ^con^ -- list of potential continuous confounding variables. 
    ^cat^ -- list of potential categorical (nominal) confouding variables. 
    ^model(logit^|^poission^|^cox)^ -- specifies the regression method, default is 
        logistic regression (^logit^), ^poisson^ and ^cox^ can be 
        specified. 

    ^noggraph^ -- suppresses graphs. By default, a graph is given.   

    ^backward^ -- specifies selection strategy as "back deletion method", 
        The default is the "forward selection" method.

    ^expos(^varname^)^ -- specifies a variable that reflects the amount of 
        exposure over which the ^yvar^ events were observed for each 
        observation. This is the option only for ^poisson^ regression.      

    ^dead(^varname^)^ -- specifies the name of a variable recording 0 if 
        censored and nonzero -- typically 1 -- if failure. If dead() is 
        not specified, all observations are assumed to have failed.  This
        is the option only for ^cox^ regression.

    ^coef^ -- graphs regression coefficients instead of effect estimates 
        for graphs.

    ^detail^ -- gives details at each step. The default is a summary. 

    ^level(^#^)^ -- specifies the confidence level, in percent, for 
        calculation of confidence intervals (default=95%)      

    ^graph_options^ -- some options with ^graph^ are allowed; see help 
        @graph@.


Examples
--------

 .  ^epiconf dead ab_uria, con(age weight) cat(smoke hypert sex) yline(1)^
        Calculates and graphs crude and adjusted odds ratios using
        logistic regressions with a forward selection approach. A reference
        line of non-effect (OR=1) is given.         
 .  ^epiconf dead ab_uria, con(age weight) cat(sex hypert) m(poisson) ^
    ^e(time) b detail^
        Calculates and graphs crude and adjusted rate ratios using poisson
        regressions with a backward deletion method. Details at each step 
        are given.

 .  ^epiconf time ab_uria, con(age weight) cat(sex hypert smoke) m(cox)^
    ^dead(dead) b nograph^
        Calculates crude and adjusted hazard ratio using Cox proportional 
        hazard models with a backward deletion method, suppressing graph.


Author
------

    Zhiqiang Wang
    Menzies School of Health Research
    Darwin, Australia
    wang@@menzies.su.edu.au


Also see
--------

     STB: STB-49 sbe27