Extract model's p-value from fitnlm

Dear all,

I am using the Statistical Toolbox (Matlab2013b) to perform a regression analysis. Using apply different regression functions to my dataset (linear,quadratic,exponential) using "fitnlm". I want to choose the best function based on R^2 and p-value. The R^2 is found in the model's output fields. The overall p-value I can't find though it is displayed in the command window: Nonlinear regression model, e.g.:

    y ~ b1*x1^2 + b2*x1 + b3

Estimated Coefficients:
          Estimate    SE          tStat       pValue  
    b1     0.02495    0.010583      2.3575    0.027708
    b2    -0.75062      1.0529    -0.71294     0.48337
    b3      14.633      24.854     0.58877     0.56201

Number of observations: 25, Error degrees of freedom: 22 Root Mean Squared Error: 8.27 R-Squared: 0.897, Adjusted R-Squared 0.887 F-statistic vs. constant model: 95.5, p-value = 1.43e-11

My tool should extract the model p-value = 1.43e-11 automatically (not the individual coefficients' pValue in the table). How can I do this?

Thanks!

Answer 1

[Bennett Landman]

In 2016a, this information is accessible via: mdl.Coefficients.pValue (assuming that mdl is the structure returned by fitnlm).

Answer 2

[dpb]

My release doesn't include fitnlm so can't test but doc page indicates that you can recompute the p-values of the various statistics via the coeftest method although the value doesn't seem to be returned as a property.

[Steven]

Correct, the coefTest function calculates the p-value as well. The outcome is slightly different though. Probably a different approach is used. I think I'll just use the coefTest.

Answer 3

[the cyclist]

It seems to me from the documentation that the p-value is not an output.

The F-test calculation is in the file NonLinearModel.m. [Lines 800-834 is the ftest() function in R2014a.] I haven't checked, but it looks like you might be able to reconstruct the p-value from the other outputs.

[Steven]

Thank you for the help! How can I open this NonLinearModel.m file?

[the cyclist]

Type

edit NonLinearModel

at the command prompt

(Be careful not to make changes. You might want to copy that another file.)