[R] "mantel.haenszel.test for trend in S-plus doesn't work i R"

[David Winsemius]

On Apr 4, 2010, at 6:51 AM, Fredrik Lundgren wrote:

> Dear R'ers,
>
> When I used S-plus i wrote a small program for a Mantel-Haenszel test
> for trend (I think it worked). Unfortunately I can't get it working in
> R.
> It appears as if my use of 'el' is the problem but I can't sort it  
> out.
>
> Error in apply(array, c(, 2, 3), function(el) el * 1:s) :
>   argument is missing, with no default
>
> Further down in the program I use 'el' as well



Don't think it has anything do do with "el". In this line:

"  O <- sum(apply(apply(array, c(, 2, 3), function(el) el * 1:s)[,
+ 1,  ], 2, sum))"

... try removing the extraneous "," in the dimension spec of the inner  
apply call. I cannot be sure that the logic is correct but at least it  
allows the function to perform calculation and make a reprot.

(I also removed the quotes around the function name since that seemed  
very un-R-like.)

It might require some re-arrangements of the data and changing the age  
variable to a numeric score, but you considered using:

glm( c(Case, Control) ~ age + sex, family="binomial")

-- 
David
>
> Can this be sorted out?
>
> Fredrik
>
>
>
> #######
> "mantel.ext.test" <-
> function(array)
> {
> # Mantel-Extension Test  -- Testing for trend in the presence of
> # confounding ref. Rosner : Fundamentals of Biostatistics, 5th
> # ed, page 606-609
> # array is an 3-dim array with as many strata (last dimension)
> # you like case-control as second dimension and the dimension
> # for which a trend should be tested as the first dimension
> # Example:
> # snoring <- array(c(196, 223, 103, 603, 486, 232, 188, 313,
> # 232, 348, 383, 206), c(3,2,2))
> # dimnames(snoring) <- list(age = c('30-39', '40-49', '50-60'), #
> status = c('Case', 'Control'), sex = c('Women', 'Men'))
> #
> # mantel.ext.test(snoring)
> #   Data: snoring
> # , , Women
> #       Case Control
> # 30-39  196     603
> # 40-49  223     486
> # 50-60  103     232
> # , , Men
> #       Case Control
> # 30-39  188     348
> # 40-49  313     383
> # 50-60  232     206
> # The trend of snoring with age , controlling for sex , has a
> # Mantel-Extension-test chi-square = 35.06 (df = 1 ) p-value = 0
> #  Effect is " increasing " with increasing age
>     s <- dim(array)[1]
>     O <- sum(apply(apply(array, c(, 2, 3), function(el) el * 1:s)[,
> 1,  ], 2, sum))
>     tot <- apply(array, c(3), sum)
>     sum.case <- apply(array, c(2, 3), sum)[1,  ]
>     E <- sum((apply(apply(apply(array, c(1, 3), sum), 2, function(
>         el) el * 1:s), 2, sum) * sum.case)/tot)
>     s1 <- apply(apply(apply(array, c(1, 3), sum), 2, function(el)
>     el * 1:s), 2, sum, simplify = F)
>     s2 <- apply(apply(apply(array, c(1, 3), sum), 2, function(el)
>     el * (1:s)^2), 2, sum)
>     V <- sum((sum.case * (tot - sum.case) * (tot * s2 - s1^2))/(tot^
>         2 * (tot - 1)))
>     A <- abs(O - E) - 0.5
>     chi2 <- (abs(O - E) - 0.5)^2/V
>     p <- 1 - pchisq(chi2, 1)
>     incr <- ifelse(A > 0, "increasing", "decreasing")
>     cat("\n", "Data:", deparse(substitute(array)), "\n")
>     print(array)
>     cat("\n", "The trend of", deparse(substitute(array)), "with",
>         attr(dimnames(array), "names")[1], ", controlling for",
>         attr(dimnames(array), "names")[3], ",", "has a", "\n",
>         "Mantel-Extension-test chi-square =", round(chi2, 3),
>         "(df =", 1, ")", "p-value =", round(p, 6), "\n",
>         "Effect is \"", incr, "\" with increasing", attr(
>         dimnames(array), "names")[1], "\n")
> }
>
> ######
>
>
> ########################
>
> Fredrik Lundgren
> fredrik.bg.lundgren at gmail.com
>
> Engelbrektsgatan 31
> 582 21 Linköping
> tel		013 - 47 30 117
> mob 	0706 - 86 39 29
>
> Sommarhus: Ljungnäs 158
> 380 30 Rockneby
> 0480 - 650 98
>
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

David Winsemius, MD
West Hartford, CT