Test if coefficient is different from 1

Hi All,

I'm using fitnlm function, and would like to test if one of the coefficients is not equal to 1. Testing if it is different from 0 is easy (coeftest), but this function doesn't seem to have the ability to specify a null hypothesis other than zero. Any thoughts on how I can test for a difference between my coefficient (lambda) and 1?

Thanks,

Steven

Answer 1

[Star Strider]

The simplest way is to see if the confidence interval for that coefficient (lambda) includes 1. If it does not, the value of the coefficient it is significantly different from 1 at the given probability. (This is the same as determining if it is different from zero.)

Rather than using fitnlm, it will probably be easier to run your regression again with nlinfit, and then use nlparci to calculate the confidence intervals.

[Steven Burnett]

Great answer! Thanks

[Star Strider]

My pleasure!