pint.utils.FTest

pint.utils.FTest(chi2_1, dof_1, chi2_2, dof_2)[source]

Run F-test.

Compute an F-test to see if a model with extra parameters is significant compared to a simpler model. The input values are the (non-reduced) chi^2 values and the numbers of DOF for ‘1’ the original model and ‘2’ for the new model (with more fit params). The probability is computed exactly like Sherpa’s F-test routine (in Ciao) and is also described in the Wikipedia article on the F-test: http://en.wikipedia.org/wiki/F-test The returned value is the probability that the improvement in chi2 is due to chance (i.e. a low probability means that the new fit is quantitatively better, while a value near 1 means that the new model should likely be rejected).

Parameters:

  • chi2_1 (float) – Chi-squared value of model with fewer parameters

  • dof_1 (int) – Degrees of freedom of model with fewer parameters

  • chi2_2 (float) – Chi-squared value of model with more parameters

  • dof_2 (int) – Degrees of freedom of model with more parameters

Returns:

ft – F-test significance value for the model with the larger number of components over the other.

Return type:

float