mdl object: put "F-statistic vs. constant model: ... , p-value = ..." into variable

fitlm() displays, among other things, the F-statistic and p-value for the fit as a whole. I can't find these values in the mdl object, however. Could someone point me to how I can find, or calculate this value using the data in the mdl object?

Answer 1

[gusjammo]

Hello Evan, I think that you may be able to do it with coefTest(mdl), although this defaults to a horizontal line along the x-axis, so you might have to override the default with the average y-value, for instance.

[fateme]

Hi the question is for long time ago but maybe this answer helps others. Yes as Gusjammo said based on the help of Matlab this works:

[p,F] = coefTest(mdl)

However for me,this function gives NaN inspite of showing values of p , F by typing 'mdl'.

Answer 2

[the cyclist]

I'm not sure it is very well documented, but if you type

mdl.

at the command line, you will see the list what is included in the model object.

The P-values are

mdl.Coefficients.pValue

or

mdl.anova.pValue

and the F-stat is

mdl.anova.F

[the cyclist]

Oh, hm. That does not seem to be the same F-stat as in the simple display. I feel like it must be in the model object, but I poked around some more and did not find it.

[Evan]

Are you sure that's right? All of those fields have at least N values, where N is the number of independent variables. I'm looking for a single test statistic.

[Evan]

It is curious though that anova seems to be a property of the mdl object, yet it's not listed when you call the function properties(mdl)... Typing "mdl." in the command line for me just returns an error: "Error: Expression or statement is incomplete or incorrect."

[the cyclist]

I misspoke when I said to enter "mdl." in the command window. What you should do is type "mdl." without hitting Enter, and then hit Tab to see the possible completions of that. Then you will see things like Coefficients, etc.