0

Apparently older versions of R provide some additional info. about max. "ncp" (non-centrality parameter) in ?df. Does anyone know what the max. value for ncp under df() (i.e., F distribution) supported by R is?

P.S. I know for dt() (i.e., t distribution), max. ncp allowed in R is 37.62.

rnorouzian
  • 7,397
  • 5
  • 27
  • 72
  • I am not sure about a "maximum" value. You may take a look at [this question](http://stackoverflow.com/questions/39183938/rs-t-distribution-says-full-precision-may-not-have-been-achieved/39184616#39184616) for an example with the t distribution where the value is much lower than the maximum you mention as well as the documentation in `?Chisquare`. – lmo Jan 23 '17 at 16:45
  • @Imo, for `?Chisquare` in my R it vaguely says it supports `ncp < 80`. But for `?df`, R doesn't even say that. – rnorouzian Jan 23 '17 at 17:12
  • ?df says: "The non-central F distribution is again the ratio of mean squares of independent normals of unit variance, but those in the numerator are allowed to have non-zero means and ncp is the sum of squares of the means. See Chisquare for further details on non-central distributions." I guess this gives some idea of the formula. In the notes section, it vaguely or cautiously says; "The code for non-zero ncp is principally intended to be used for *moderate* values of ncp: it will not be highly accurate, especially in the tails, for large values." (using R 3.2.5) – lmo Jan 23 '17 at 17:16
  • Right, I saw that, well, there are ways to get the `ncp` for F dist. using (SSeffect/ SSerror) * N [some times `df denom.`. istead of `N`), but I'm writing an `optimize()` function in which upper and lower limits of `ncp` are important. – rnorouzian Jan 23 '17 at 17:23

0 Answers0