Re: Probability scale

On Fri, 26 Sep 2003 19:45:10 -0500
"John W. Eaton" <[email protected]> wrote:

> On 26-Sep-2003, Phi Reardon <[email protected]> wrote:
> 
> | It seems to me (correct me if I am wrong) that octave can plot on
> | linear or log scales, but not probability scales.  This would not be
> | too hard to implement, it seems to me, but is not a current octave
> | (or gnuplot) capability.  If it were in place, then log-probability
> | graphs of distributions like the log-normal distribution would be
> | easy.  Since this is a common sort of distribution, and even a
> | common sort of graph paper, how about it?
> 
> OK, write it up and send in a patch.
> 
> If there is something similar in Matlab, it would probably help to
> have a compatible interface.
> 
> jwe
> 

A patch is probably outside my skillset, but here is a little routine that locates the tic marks.  I hope it will inspire someone else to look at implementing it in octave/gnuplot.


function ans = pinv(y)
# probinv calculates the inverse of the normal cumulative distribution function
  ans = erfinv(y/50.0 - 1.0);
#  printf("y=%f,   ans=%e\n",y,ans);
endfunction
                                                                                
tics = [0.1;0.2;0.3;0.4;0.5;0.6;0.7;0.8;0.9;1.0;2.0;3.0;4.0;5.0;6.0;7.0;8.0;9.0;10.0;20.0;30.0;40.0;50.0;60.0;70.0;80.0;90.0;95.0;99.0;99.9];
                                                                                
L=15.0  #length of the scale in some unit, e.g. cm.
                                                                                
jmax=length(tics);
for j = 1:jmax
# v evaluates to 0 for j=1, L for j=jmax.
# It represents the distance from the base of the scale to the jth tic
  v(j) = L*(pinv(tics(1)) - pinv(tics(j)))/(pinv(tics(1)) - pinv(tics(jmax)));
  printf("  %f     %e\n",tics(j), v(j));
endfor