One-Sample Anderson-Darling Test Table

The following tables give the critical values for the one-sample Anderson-Darling test for the specified distributions, as described in One-Sample Anderson-Darling Test.

Critical values table 1

Critical values table 2

Laplace distribution table

Generalized Pareto critical values

Critical values: Cauchy distribution

Benford critical values

α represents the significance level and n = the sample size.

Note that k in the Gamma table is the distribution’s shape parameter. The symbol > in that table represents a shape parameter > 20. The symbol > in the Laplace or Cauchy table represents a sample size > 100.

Download Tables

Click here to download the Excel workbook with the above tables.

References

D’Agostino, R. B., Stephens, M. A. (1986) Goodness-of-fit techniques. Marcel Dekker, Inc.
https://www.hep.uniovi.es/sscruz/Goodness-of-Fit-Techniques.pdf

Stephens, M.A. (1979) The Anderson-Darling statistic. Technical Report No. 39. Stanford University.
https://apps.dtic.mil/sti/tr/pdf/ADA079807.pdf

Puig, P. and Stephens, M. A. (2000) Tests of fit for the Laplace distribution, with applications
https://nova.newcastle.edu.au/vital/access/services/Download/uon:5419/ATTACHMENT03?view=true

Arshad, M., Rasool, M. T., Ahmad, M. I., (2003) Anderson Darling and Modified Anderson Darling tests for Generalized Pareto distribution
https://scialert.net/abstract/?doi=jas.2003.85.88

36 thoughts on “One-Sample Anderson-Darling Test Table”

  1. Dear Charles,
    Is there any source to find the critical values of Anderson-Darling test for two shape parameter Beta distribution?
    I will thank you in advance for your reply.

    Reply
  2. Dear Charles,
    I would like to ask you where can i find the critical values of Anderson-Darling for Beta distribution.
    Is there any source that i could find them?
    I will thank you in advance for your reply.

    Reply
    • Hello Paul,
      A random variable x is log-normally distributed provided the natural log of x, ln x, is normally distributed.
      Thus if you have data x1, x2, …, xn that comes from a log-normal distribution, then ln x1, ln x2, …, ln xn is normally distributed. Thus, you can use the critical values of the normal distribution.
      Charles

      Reply
    • Hi Paul,
      See the following articles:
      Jantschi and Bolboaca (2018) Computation of Probability Associated with Anderson–Darling Statistic
      Shin et al. (2012) Assessment of modified Anderson-Darling test statistics for the generalized extreme value and generalized logistic distributions
      Charles

      Reply
  3. Hi. Might there be a small error in the values for the Lognormal d row? Seems like the values for 0.01 and 0.025 cannot be correct.

    Reply
    • Rafael,
      If the uniform distribution is over the interval (0,1), then the critical values are the same as for the generic case (in fact it is the generic case).
      Charles

      Reply
      • Dear Charles,

        Thanks for the answer. Sorry it bothers you but how should I do if the uniform distribution has limits other than 0 and 1?

        Thanks again

        Reply
        • Hi Rafael,
          If the limits are known, say (a,b), then you can map each value x in this range to a + (x-a)/(b-a). This maps x to a value in the interval (0,1). You can then use the critical values for (0,1).
          If the values a and b are not known, then the situation is more complicated and I haven’t yet explored how to use Ander-Darling in this case.
          Charles

          Reply
  4. Dear Charles,

    I would like to know the formula for testing such critical values for different distribution

    Kind regards.

    Reply
  5. good afternoon mr
    i wonder to ask about shape value
    is the different between shapre value and shape parameter for gamma distribution?

    Reply
  6. Excuse me mr charles
    I want to ask something
    The critical value of anderson darling, is just for 5 distributions above?what about rayleigh distribution sir?
    Thank you

    Reply
    • Izza,
      I don’t have critical values for the Rayleigh distribution.This distribution is the gamma distribution with alpha = 1 and beta = 2. The table contains critical values for this distribution with an unknown shape parameter, but not with a known shape value.
      Charles

      Reply

Leave a Comment