your instance of java.util.Random inside the static method randomInt
in your gatling class.
Basically, every time you run randomInt you're constructing a new
Random object but with no seed. Java uses a default seed that, I
think, is something like a millisecond timestamp. BUT since in a
simulation like this, you're running randomInt multiple times per
milisecond, it means that Random is getting seeded with the same value
multiple times in a row.
To avoid this you should really only use one instance of Random for
any particular of the program. So your gatling class should be
something like:
public class gatling{
// instance of Random to use whereever you need it
// will be constructed the first time a static method is called on
the gatling class, and then not again.
private static Random myRand = new Random();
public static int randomInt(int lo, int hi){
return myRand.nextInt(....)
//etc.
}
public static char randomChar(){
return (char)( myRand.nextInt(26))+'a'
}
//etc.
}
--------------------------------------------------------------
Baker Franke
Computer Science Dept.
The University of Chicago Laboratory Schools
773.702.5419
On Oct 5, 2009, at 3:13 PM, Keith E Gatling wrote:
> Last week my class discussed the statistical question of what the
> likelihood
> was of a family with two kids having both of them be girls. We knew
> that it
> was supposed to be 25%, but the programs that my students wrote all
> came up
> with likelihoods in the area of of 40%, and this was after multiple
> runs
> with sets of 100, 1000, and even 30,000,000 families.
>
> I decided to find out what they were doing wrong by writing the code
> myself
> (without looking at theirs) and checking for the correct proporotion
> of boys
> and girls in each column. It worked like a charm - except that even
> I got
> the skewed stats of roughly 40% for two-girl families. In fact, I was
> getting roughly 40% for families with both two girls and two boys,
> while
> families with one of each were roughly 20%
>
> I checked this out with my sister the statistician, who said that my
> logic
> was correct, and then wrote it herself in SAS. She got the expected
> figures.
> But we're all still a bit confused as to why we keep getting the
> right stats
> for each column (Kid A and Kid B), but totally wrong stats for BB,
> BG, and
> GG.
>
> Do any of you Java experts out there want to take a look at my code to
> either try to figure out what we're doing wrong, or maybe run it in a
> different environment to see if you get different results? We've
> been doing
> it in BlueJ on Macs. Do you get different results in BlueJ in
> Windows or
> Linux? Do you get different results in a different programming
> environment?
>
> The main code in question is at:
> http://www.gatling.us/keith/class/javajive/TwoKids.java
> The gatling class it refers to for gatling.randomInt is at
> http://www.gatling.us/keith/class/gatling.java
> For what it's worth, the gatling.randomInt method is merely an
> interface to
> the standard Java method of creating random numbers.
>
> Can anyone solve this conundrum for us?
>
> Thanks!
> --
>
> keg
>
> ========================================
> Keith E Gatling
> Email: keith@gatling.us
> Blog: http://wordfromg.blogspot.com
> Website: http://www.gatling.us/keith
> The fact that I'm open-minded doesn't mean that I have to agree with
> you.
> ========================================
>
> [ For info on ISED-L see http://www.gds.org/ISED-L ]
> Submissions to ISED-L are released under a creative commons,
> attribution, non-commercial, share-alike license.
> RSS Feed, http://listserv.syr.edu/scripts/wa.exe?RSS&L=ISED-L
>
[ For info on ISED-L see http://www.gds.org/ISED-L ]
Submissions to ISED-L are released under a creative commons, attribution, non-commercial, share-alike license.
RSS Feed, http://listserv.syr.edu/scripts/wa.exe?RSS&L=ISED-L
