The Confidence Quantifier

[Eduardo]

[BillyJoe]

I thought I was bad, sometimes, but you guys are something else.. in a nice way...  



 
Frank, I wish you good luck with your programming efforts..

[Frank Kneeland]

OK I rechecked the Chi equations by inputting some CLEARLY NON-RANDOM DATA, and surprise surprise...it did indeed tell me the data was no beuno.The problem would seem to be an issue of sensitivity. It's just not very sensitive and it goes from telling you your data is fine to disavowing it far faster than I expected.I know you are familiar with the chi results. It hovers around .99999 and then plummets like a stone to near zero with only slight alterations in the data.Fine...fine...fine...no good!!!That's not what I was expecting. I was expecting a slow gradual drop from 1 to 0. It gave me .99 results with data I made up. Yes it did rate the truly random data as more likely to be random, but the difference between .9999 and .99 is lost on most people.It also seems to be more sensitive to overages than underages.I have no idea if this is normal or if I'm doing something wrong. This makes it hard to give the utility a pass or fail when I don't know what a pass or fail would look like.To me it seems woefully inappropriate to the task I hoped to use it for.Any thoughts???

[Eduardo]

No bueno?
Issue of sensitivity?
Underages?

Can we go back to talking about your program instead of my prom?

[faygo]

No bueno?
Issue of sensitivity?
Underages?

Can we go back to talking about your program instead of my prom?
 Can only imagine how traumatic your prom experience must have been for  it is so ingrained in your memory.The unfortunate thing is it made me recall mine.  By all means, back to the program Frank.

[BillyJoe]

Not sure this is the opinion that you seek, Frank, but to me, it sounds like it is doing what it is designed to do, trying to prove something to be non-random. It either is PROVED NON-RANDOM, or not. If it cannot be PROVED NON-RANDOM, then it gives it the benefit of the doubt, and says there is a high likelihood of randomness. Not that it ISN'T non-random, just can't prove it. Kinda like innocent until proven guilty. It does not seem that there is a 'degree' of randomness - it either is or it isn't non-random.
 
Of course, I haven't been sleeping well lately, so this may all be hogwash..    

[Frank Kneeland]

Not sure this is the opinion that you seek, Frank, but to me, it sounds like it is doing what it is designed to do, trying to prove something to be non-random. It either is PROVED NON-RANDOM, or not. If it cannot be PROVED NON-RANDOM, then it gives it the benefit of the doubt, and says there is a high likelihood of randomness. Not that it ISN'T non-random, just can't prove it. Kinda like innocent until proven guilty. It does not seem that there is a 'degree' of randomness - it either is or it isn't non-random.
 
Of course, I haven't been sleeping well lately, so this may all be hogwash..    Well if you are right, I may be able to release it very soon. I'd like the math boys to chime in first.

[Frank Kneeland]

I'll be out of the house until Sunday, but come Monday I'd like to resolve this so I can decide on whether to scrap the current utility or post it.~FK

[shadowman]

Frank, if you can find the break-point where it drops like a stone then all you need to do is normalize the higher values to a range that you think gives a better feeling of the randomness. For example, if .99 is the lower limit then map that into 1 and every value above that an increasing value. Set anything below that to zero (or non-random) and you have the range you can work with.
 
Something like Y=(X-.99)*100 followed by if (Y<0) then Y=0.

[Frank Kneeland]

Frank, if you can find the break-point where it drops like a stone then all you need to do is normalize the higher values to a range that you think gives a better feeling of the randomness. For example, if .99 is the lower limit then map that into 1 and every value above that an increasing value. Set anything below that to zero (or non-random) and you have the range you can work with.
 
Something like Y=(X-.99)*100 followed by if (Y<0) then Y=0.Received and thinking.