[R] prop.trend.test issue

[Michael Dewey]

At 05:55 03/12/2006, Ethan Johnsons wrote:
>I have the clinical study data.
>
>                                            Year 0       Year 3
>Retinol (nmol/L)        N       Mean +-sd       Mean +-sd
>Vitamin A group         73      1.89+-0.36      2.06+-0.53
>Trace group                57      1.83+-0.31     1.78+-0.30
>
>where N is the number of male for the clinical study.
>
>I want to test if the mean serum retinol has increased over 3 years
>among subjects in the vitamin A group.

If you want to test means why did you think a test for proportions 
was a good idea?


>>  1.89+0.36
>[1] 2.25
>>1.89-0.36
>[1] 1.53
>>2.06+0.53
>[1] 2.59
>>2.06-0.53
>[1] 1.53
>
>
>>prop.trend.test(c(2.25, 1.53),c( 2.59, 1.53))
>
>        Chi-squared Test for Trend in Proportions
>
>data:  c(2.25, 1.53) out of c(2.59, 1.53) ,
>using scores: 1 2
>X-squared = 0.2189, df = 1, p-value = 0.6399
>
>The issue I am seeing that N of Vitamin A group = 73 seems not reflected.
>This leads me to think that I can't implement the test based on the
>data just presented.
>Nor a two-tailed test is possible.
>
>        2-sample test for equality of proportions with continuity correction
>
>data:  c(2.25, 1.53) out of c(2.59, 1.53)
>X-squared = 0, df = 1, p-value = 1
>alternative hypothesis: two.sided
>95 percent confidence interval:
>-0.6738203  0.4112720
>sample estimates:
>   prop 1    prop 2
>0.8687259 1.0000000
>
>Warning message:
>Chi-squared approximation may be incorrect in: prop.test(c(2.25,
>1.53), c(2.59, 1.53))
>
>Any ideas, please?
>
>thx
>
>ej
>
>

Michael Dewey
http://www.aghmed.fsnet.co.uk