[R] mult comp significance

Spencer Graves spencer.graves at pdf.com
Thu Aug 3 13:07:10 CEST 2006


	  I read two questions:  physical meaning and how to display only 
significant differences.


PHYSICAL MEANING

	  A p value is the probability of obtaining by chance alone a result at 
least as extreme as what we observe.  A "simultaneous" or "adjusted" p 
value is the probability that at least one of the multiple tests was at 
least as extreme as what observed.

	  A (1-alpha) confidence interval or region is a random set that 
includes the true but unknown value with probability at least (1-alpha). 
  A set of (1-alpha) simultaneous confidence intervals is a random set 
in the multidimensional space that includes the true but unknown values 
of all parameter comparisons with probability at least (1-alpha).


SELECTING ONLY SIGNIFICANT DIFFERENCES

	  Consider the following modification of the first example in the 
"simint" help page:

library(multcomp)
data(recovery)
RecInts <- simint(minutes~blanket, data=recovery, type="Dunnett",
                     conf.level=0.9, alternative="less",eps=0.0001)

	  To see the structure of "RecInts", look at "str(RecInts)".  This 
indicates that "RecInts" is a list, and one of its components is a named 
vector called "p.value.adj".  This suggests we try the following:

 > with(RecInts, p.value.adj[p.value.adj<0.05])
blanketb2-blanketb0
        5.484862e-05

	  Hope this helps.
	  Spencer Graves

Nair, Murlidharan T wrote:
> This has a stats question and a R question.  I am sure there are many
> core statisticians here how would know the answer to this simple
> question. In determining the significant comparisons using the methods
> in multcomp, the ones that are designated as significant are the ones
> that do not intersect the zero line. What is the physical meaning of
> this and why are those considered significant? I can sort those out and
> pick out by their adjusted pvalues to pick the top ones correct? Is
> there an method the multcomp that will output only the significant ones?
> 
> Thanks ../Murli
> 
>    
> 
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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.



More information about the R-help mailing list