10-28-2016 11:51 AM
In my experience, it is a good idea to believe Darin when it comes to algorithm, so my main point is that the help needs to be corrected and there should be a CAR. OTOH, there is a nonzero probability that NI could change the algorithm in the future to conform to the wording in the help. In general, I think it is a good idea to do whatever Ravens and me have been doing and code according to the help, just to be safe. 😄
10-28-2016 11:55 AM
10-28-2016 12:05 PM
@Norbert_B wrote:The second is to accept that 1000 iterations maybe represent a trend, but statistically speaking, is an insignificant amount of test iterations. For such small tests, i would recommend to iterate for more than 1M times (at least) in order to get representative data.
1000 is plenty to demonstrate a trend. Statistics gives us all the tools to assign confidence limits and even with 1000 tests we can say if the outcome is fair or not within a given confidence, especially if the difference is so large.
09-10-2019 07:31 PM - edited 09-10-2019 07:38 PM
That is why I seem to prefer the riffled index out of riffle vi for uniform distribution of integer values from a discrete number of values.
Well, the new and improved riffle vi...you can't argue with statistics too much!
At Norbert.... then why the two poly instances of generate white noise^? One using a prime / coprime method and the other depending on an irrational number? (BTW, one too may "e"s in xxxMYER in the help file)