How can I get the pval working with percentages considering a table with NA values and data essentially constant (100% or 0%?

Asked May 02 '20 at 14:23

Active May 02 '20 at 15:42

Viewed 36 times

I am trying to get the pval from a bigdata table (700 observations and 6 variables). I can't calculate the p values due to repeated values of some percetanges (100% or 0%)

The table more or less represent my doubt. What can I do to get my p values at least in the rows that do not display repeated values?

b<-cbind(c(100,50,20),c(100,49,19),c(100,48,18),c(100,47,17), c(100,NaN, 16))

pval<-apply(b, 1, function(x) t.test(x[1:2], x[3:4], alternative = "two.sided", var.equal = T))

Error in t.test.default(x[1:2], x[3:4], alternative = "two.sided", var.equal = T) : data are essentially constant

edited May 02 '20 at 15:42

asked May 02 '20 at 14:23

Eros Marin Millan

i guess you should not do a t.test on percentages? In any case, using the example you provide, i don't get an error. you can wrap the t.test around try() – StupidWolf May 02 '20 at 14:56
Sorry I made a mistake in the data. Now is corrected – Eros Marin Millan May 02 '20 at 15:47
thanks. you tried using pval<-apply(b, 1, function(x)try(t.test(...)))? – StupidWolf May 02 '20 at 15:48
With try I get the same outcome. – Eros Marin Millan May 02 '20 at 17:01
yes.. you get an error. But you should also get a pval list. the entries that throw an error have "error.." – StupidWolf May 02 '20 at 17:42
1

Thank you! I just did that and removed the error row and now it works! Thank you! – Eros Marin Millan May 02 '20 at 18:03

How can I get the pval working with percentages considering a table with NA values and data essentially constant (100% or 0%?

0 Answers0