File Exchange

## FACTORIALRATIO

version 1.4 (2.81 KB) by

Ratio of factorials, as in (a!b!c!...) / (d!e!f!g!...) (version 2.0, oct 2012)

Updated

FACTORIALRATIO - ratio of factorial numbers

FR = FACTORIALRATIO (A,B) calculates the ratio between products of
factorials specified by A and B. A and B are lists of positive
integers, with N and M elements, respectively. The number of values in
each list may differ. The factorial ratio is based on the following formula:

(A1! * A2! * ... * AN!)
FR = -------------------------
(B1! * B2! * ... * BM!)

However, an approach is taken that increases the accuracy when A or B
contain larger values, as mentioned in the help of FACTORIAL.

Examples:

f1 = factorialratio([10 8 6],[9 6 5]) ;
f2 = 10 * 8 * 7 * 6 ; % by hand
f3 = (factorial(10) * factorial(8) * factorial(6)) ./ ...
(factorial(9) * factorial(6) * factorial(5)) ;
disp('f1, f2, f3 = ') ; disp([f1 f2 f3])

% Also for larger values FACTORIALRATIO works correctly
f1 = factorialratio(150,143) ;
f2 = prod(144:150) ; % by hand
% where FACTORIAL lacks precision
f3 = factorial(150) ./ factorial(143) ;
disp('f1, f2, f3 = ') ; disp([f1 f2 f3])
disp(' Differences with f2 = ') ; disp([f1-f2 f2-f2 f3-f2])

% and even for extreme numbers things turn out well
f1 = factorialratio(30000,29998) ;
f2 = (30000 * 29999) ; % by hand
% whereas calls to FACTORIAL fail miserably ...
f3 = factorial(30000) ./ factorial(29998) ;
disp('f1, f2, f3 = ') ; disp([f1 f2 f3])

Note that, for instance, factorialratio(200,1) also fails, of course.
For such occasions other engines have to be used.

Jos (10584)

### Jos (10584) (view profile)

Version 2.0 (oct 2012) fixed the bug mentioned by Charles.

Charles

### Charles (view profile)

Version 1.1 (feb 2009) of this code has a bug.

x(1)=factorialratio(2,1); %should be 2
x(2)=factorialratio(1,2) %should be .5

x =

1 1

John D'Errico

### John D'Errico (view profile)

Sorry, I forgot to point out what I feel to be one of the best features of this code - the copious internal comments that explain how it works.

I've often argued that the best MATLAB code should aim for a 1-1 ratio of comments to lines of code. That is, each line of code should have an associated comment line, at least in those places where the code is doing something tricky. This code goes beyond my aim. People interested in how to write good comments should look at what the author did here and learn from it.

John D'Errico

### John D'Errico (view profile)

Good help as you would expect from Jos. Well written. An interesting solution for the computation itself.