[R] About continuity correction option in the mantelhaen.test function.

[Peter Dalgaard]

Paek, Insu wrote:
> Hello,
>    I was using the mantelhaen.test function (2x2 J tables for
> conditional independence testing). I noticed that the option for the
> continuity correction (correction=T or correction=F) sometimes made a
> difference while sometimes it did give the same results regardless of
> correction=T or correction=F.
> 
>   Does the continuity correction apply only in certain cases or in
> certain strata with some rules ?
>   Could you let me know why I am seeing no difference sometimes
> regardless of correction=T or F ?

An example would help...

According to the code, the continuity correction is turned off if DELTA 
< 0.5, so you should see it only in cases where the p-value is close to 
1. However, the output should say so:

             DELTA <- abs(sum(x[1, 1, ] - s.x[1, ] * s.y[1, ]/n))
             YATES <- ifelse(correct && (DELTA >= 0.5), 0.5, 0)
             STATISTIC <- ((DELTA - YATES)^2/sum(apply(rbind(s.x,
                 s.y), 2, prod)/(n^2 * (n - 1))))
....
             METHOD <- paste("Mantel-Haenszel chi-squared test",
                 ifelse(YATES, "with", "without"),
                               "continuity correction")

What worries me more is that it seems that 'alternative' is being 
ignored in the !exact case:


 >  mantelhaen.test(UCBAdmissions,alt="l")$p.value
[1] 0.2322635
 >  mantelhaen.test(UCBAdmissions)$p.value
[1] 0.2322635
 >  mantelhaen.test(UCBAdmissions,exact=T)$p.value
[1] 0.2277625
 >  mantelhaen.test(UCBAdmissions,exact=T, alt="l")$p.value
[1] 0.1159937



-- 
    O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
   c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
  (*) \(*) -- University of Copenhagen   Denmark      Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)              FAX: (+45) 35327907