A Using Statistics to Test for Normality of Pi

AI Thread Summary
Testing the normality of π involves assessing whether each digit appears with equal frequency, and one suggested method is to randomly sample strings of size 20 and analyze the frequencies using a chi-squared test. This approach is deemed sound for evaluating uniform distribution among the digits. There is a question about whether normality is dependent on the numeral base, as patterns may emerge in one base that appear random in another. Research indicates that π has shown normality up to 22.5 trillion decimal places in base 10 and an equivalent amount in hexadecimal. The discussion highlights the importance of statistical methods in understanding the distribution of digits in π.
WWGD
Science Advisor
Homework Helper
Messages
7,692
Reaction score
12,656
Is there a " reasonable" way to test for the normality of ##\pi## , i .e., that every digit occurs with the same frequency? Someone suggested randomly sampling strings of size 20 and outputting the frequency. Then I guess we could average the frequencies among samples , use a chi-squared test. Assuming random sampling can be done, is this a sound way of testing?
 
Last edited by a moderator:
Physics news on Phys.org
Yes. The Chi-squared goodness of fit test can be used to test if the digits are from a uniform distribution.
 
Is normality dependent upon the base - is it possible for there to be a pattern in one base that then appears random in another?

anyway, here is one method that showed normality out to 22.5 trillion base 10 decimal places and an equivalent number of hexideximal ones

https://arxiv.org/pdf/1612.00489.pdf
 

Similar threads

Replies
7
Views
2K
Replies
30
Views
2K
Replies
27
Views
3K
Replies
20
Views
3K
Replies
20
Views
4K
Replies
1
Views
3K
Back
Top