I'm trying to code the NIST test from randomness
(csrc.nist.gov/publications/nistpubs/800-22-rev1a/SP800-22rev1a.pdf)
using Matlab. I wish to release the test suit to the public once completed.
However the result of the example given in the "Discrete Fourier Transform (Spectral) Test" (section 2.6 page 34) does not tally with the results of my program. I think the problem is with the DFT.
Can anyone help me to locate the fault? I'm uploading my matlab file and the reference file.
Here is the reference pdf (For some reason I cant attach it to above post)
http://csrc.nist.gov/publications/nistpubs/800-22-rev1a/SP800-22rev1a.pdf
basically, if you convert the line for N1 calculation as:
N1 =sum(peaks < T & peaks > 0);
you would have the correct value
Thanks Nevrez,
Unfortunately that modification do not bring the correct answer to the second example given below it. Calculated N1 is 48 while the reference says it should be 46. Any idea ?
I think you should check the fft data, you know half is the real data you can use, the other half basically the symmetric negative axis. Better to check which part you are using, and if the number of data is even one part will be with 1 less data.
My other guess would be some values in matlab very close to 0 but not 0 so the condition may count them as peak too.
Another think is that, maybe, some data that you are taking as peaks can be local minima instead of local maxima which refers to peaks.
I didn't check the second example but these are my possible ideas, of course they may be wrong but you may want to check. Best regards
- Badges
- Science topic