Royston approximation for Shapiro-Wilk normality test


New Member

Shapiro-Wilk article support tables only for n<=50.

Patric Royston support calculations for a(i) and P-value based on W for n<=2000. and probably ok for n<=5000
in the following article "Approximating the Shapiro-Wilk W-test for non-normality"

Are the values in Shapiro-Wilk article also approximations?

When using Shapiro-Wilk for n<=50, is it better to use Shapiro-Wilk tables or Royston approximations?
1. For a(i) calculation?
2. For P-Value calculation, based on w?