returning p-values when using lscov function

How to return the p-values of a weighted least squares regression performed using the lscov command?

Answer 1

[Tom Lane]

If you want p-values for each coefficient, then t is the coefficient estimate divided by its standard error (both returned by lscov), and the p-value is 2*tcdf(t,df), and df=length(b)-length(x) using the notation in the reference page.

[Janis]

Thank you! Is there a command for returning the r-squared, as is the case using robustfit, or must it also be calculated?

[Tom Lane]

lscov does not return R-square, so you'd have to calculate it yourself. You'd probably want to compute the various sums of squares including the weights in those formulas.

[Doug Rubino]

I'm having trouble understanding this term: 2*tcdf(t,df)

tcdf is a cumulative distribution function so it has a range of [0 1]. Doesn't this mean the p-value has a range of [0 2]?