median

[stan]

i've spent past couple of days searching for an efficient algorithm to calculate median value that would be displayed in stats tables as the average value can be misleading sometimes. however i didn't find anything that calculates median in a single iteration over a list of values. all algorithms need multiple iterations and/or store the input values in a list. that's not usable in expert lotto as it processes millions of input values.
so i was thinking that i'd allow definition of 3 to 5 intervals for each stats table and then show how many values fall into each interval. the intervals would be user-defined as the user knows best what the range of values should and what to expect.

what do you think?

[maroc]

Hi Stan,


let's try it and see

[Joe]

[quote=stan]
the intervals would be user-defined as the user knows best what the range of values should and what to expect.
what do you think?
[/quote]

This will be great stan,
Thanks

[Jerzy]

Yes, Stan, it could be useful. Probably the number of intervals would be the most important factor and, as such, it should be user-defined.

Thanks.

[CARBOB]

[quote=Jerzy]
Yes, Stan, it could be useful. Probably the number of intervals would be the most important factor and, as such, it should be user-defined.

Thanks.
[/quote]


I agree. A very useful tool.

[Midas]

This might be helpful.

Best Regards
Midas

[MartinMan]

If it is what we can get, I will take it.
It would be a Lot better than the alternative.

Martin

[Midas]

Forgot to mention, to reduce the large sample size all that is needed is to use bootstrap sampling with an acceptable confidence interval to get a nice small representative sample.

Don't forget to shuffle(randomize) the sequence prior to sampling and that's it in a nutshell a fast efficient method.

For further background details see Kardi's tutorial.

http://people.revoledu.com/kardi/tutori ... index.html

Best Regards
Midas

[CARBOB]

[quote=Midas]
Forgot to mention, to reduce the large sample size all that is needed is to use bootstrap sampling with an acceptable confidence interval to get a nice small representative sample.

Don't forget to shuffle(randomize) the sequence prior to sampling and that's it in a nutshell a fast efficient method.

For further background details see Kardi's tutorial.

http://people.revoledu.com/kardi/tutori ... index.html

Best Regards
Midas
[/quote]

Midas, thank you for the link. Very good tutorial!!!

[stan]

thanks for the link midas. i think i can implement similar algorithm that would yield acceptable results without increased memory requirements