NEWS

betaSandwich 1.0.9.9001

Add SCorSqSandwich() for the image of a signed-semipartial interval under squaring.

Added SCorSandwich(), PCorSandwich(), DeltaRSqSandwich(), and FSqSandwich(): delta-method intervals for semipartial correlations, squared partial correlations, the improvement in R-squared, and Cohen's f-squared (global and local), with normal-theory, ADF, or HC standard errors, as in betaMC and betaNB.
RSqBetaSandwich() is removed. Use RSqSandwich(), which gives the same estimates and standard errors; its intervals and tests use the normal distribution, as the other functions do, and its estimates are named rsq and adj.
The new functions and DiffBetaSandwich() return objects of class essandwich with one set of methods. DiffBetaSandwich() objects also keep the class diffbetasandwich.

Selective standardization: the arguments std_y and std_x standardize only y or only the named regressors (for example, leave a dummy variable in its 0-1 units). The defaults give the same results as before.
DiffBetaSandwich() stops when the slopes are not fully standardized.

Fixed the standard error of R-squared in RSqBetaSandwich(), which was about three times too small for every covariance type. With R-squared as a parameter, the variance of y is beta' Sigma_X beta / R-squared, so its derivatives with respect to beta and Sigma_X are now divided by R-squared.