CODEWITHSUNDEEP
python-3.xscipy
Yijiao Liu
Yijiao Liu

Reputation: 304

Why not the inverse of inverse function of a standard normal distribution calculation at scipy.stats python identical?

I tested the following example

from scipy.stats import norm
i=0.2
k=0.5
i==norm.cdf(norm.ppf(norm.cdf(norm.ppf(i)+k))-k)

The result gives False

If I check the value of norm.cdf(norm.ppf(norm.cdf(norm.ppf(i)+k))-k), it is 0.19999999999999996.

Why does this happen? Is there a better way to run the inverse of an inverse function?

Upvotes: 0

Views: 125

Answers (1)

vonbrand
vonbrand

Reputation: 11831

Because they are numerical approximations. Just like sqrt(2) * sqrt(2) is not 2. There is just no floating point number such that it's square (in floating point precision) is exactly 2.

Check out Goldberg's classic What every computer scientist should know about floating point, ACM Computing Surveys 23:1 (mar 1991), pp. 5-48.

Upvotes: 2

Related Questions