 generating random numbers between 0 and 1

ChrisE
msg:4026917  10:34 pm on Nov 17, 2009 (gmt 0)  I'm trying to write a small statistical procedure, and I need random numbers between 0 and 1 evenly distributed. rand(0,1) and mt_rand(0,1) are both causing me problems. They both seem to return either 0 or 1. I'm not sure if there is some setting my web host has so that the numbers are rounded or what. I also tried rand( 0, getrandmax()/getrandmax() ) with no luck. here is the function I'm writing: function boxmuller() { $a = mt_rand(0,1); $b = mt_rand(0,1); $bm = sqrt( 2 * log($a) ) * cos( 2 * pi() * $b); return $bm; } Any thoughts? In case anyone is curious, Box Muller is a way to generate a standard normal distribution.

pinterface
msg:4026973  12:19 am on Nov 18, 2009 (gmt 0)  Well, look at the documentation for rand [us.php.net]: int rand (int $min, int $max) 
 Note how the return value is an integer. So of course you aren't getting a float. rand(0, getrandmax()/getrandmax()) is closer (you want to divide something by getrandmax()), but think about what it does: it divides getrandmax() by itself, so is equivalent to saying rand(0, 1). To compress a random number between 0 and MAX to be between 0 and 1, you need to divide the random number itself by MAX.

ChrisE
msg:4027044  2:04 am on Nov 18, 2009 (gmt 0)  Wow, I feel silly. Thanks. I get sloppy with PHP since it isn't typed as strongly as C++.

TheMadScientist
msg:4027070  2:42 am on Nov 18, 2009 (gmt 0)  Not sure what to tell you about the 'evenly distributed' portion (you could probably accomplish it by storing rotating through a series of 10 possibilities and loop through them, EG 1000000, 1999999 ¦ 2000000, 2999999 ¦ etc.), and you can easily have control over the length of the decimal as follows: $number = mt_rand(1000000,9999999); echo $number='.'.$number; $number=floatval($number); echo "<br>"; echo gettype ($number); Not sure if it helps or not...

ChrisE
msg:4027080  3:15 am on Nov 18, 2009 (gmt 0)  Generally, the rand() functions in various languages are pretty even. Now that my Box Muller function is working, the results from the Monte Carlo simulation are spot on with what I was getting from running the sim in MS Excel. So that is, as they say, good enough for government work! Thanks for the help!


