The fift_t2z() function works as follows, on input x:

a) compute q = 1-cdf(x) = right tail probability (that F is bigger than x); this will be a number decreasing from q=1 at x=0 to q=0 as x gets big.

b) compute Q = 0.5 * q, which will range from Q=0.5 down to Q=0.

c) compute y as the non-negative value that has a normal right hand tail probability (that z > y) of Q

The effect of this is to map x in (0,infinity) to y in (0,infinity). This is basically the translation of a two-sided t-test threshold to the world of F-tests (since t**2 is an F). The alternative would be to map x in (0,infinity) to y in (-infinity,infinity). This would mean that tiny values of F (indicating a very poor fit to whatever model was used) would get a very large negative z score. If you made a color overlay of this, you would see color essentially everywhere, unless you chop off such values somehow.

If it is essential, I can provide a modified fift_t2z function (fift_t2z_1side?) to give the capability you describe.

bob cox