statistics - What's the lowest number R will present before rounding to 0? -

i'm doing statistical analysis r software (bootstrapped kolmogorov-smirnov tests) of large data sets, meaning p values incredibly small. i've bonferroni corrected large number of tests i've performed meaning alpha value small in order reject null hypothesis.

the problem is, r presents me p values of 0 in cases p value presumably small cannot presented (these large sample sizes). while can happily reject null hypothesis these tests, data publication, i'll need write p < ..... don't know lowest reportable values in r are?

i'm using ks.boot function in case matters.

any appreciated!

.machine$double.xmin gives smallest non-zero normalized floating-point number. on systems that's 2.225074e-308. however, don't believe sensible limit.

instead suggest in matching::ks.boot change line

ks.boot.pval <- bbcount/nboots

ks.boot.pval <- log(bbcount)-log(nboots) , work on log-scale.

edit:

you can use trace modify function.

step 1: @ function body, find out add additional code.

as.list(body(ks.boot)) 

you'll see element 17 ks.boot.pval <- bbcount/nboots, need add modified code directly after that.

step 2: trace function.

trace (ks.boot, quote(ks.boot.pval <- log(bbcount)-log(nboots)), at=18) 

step 3: can use ks.boot , return logarithm of bootstrap p-value ks.boot.pvalue. note cannot use summary.ks.boot since calls format.pval, not show negative values.

step 4: use untrace(ks.boot) remove modifications.