Changes and Updates for svydiags package

svydiags 0.7

• svycollinear updated to include more families: binomial and quasibinomial (logit and probit links), Gamma (inverse link), inverse.guassian (“1/mu^2” link), poisson and quasipoisson (log link). The function has also been redesigned to require the X matrix as an input, which is comparable to the requirements for svyvif. The unit variance-covariance matrix is now computed within the function based on information in the input model object (mobj).

• svyvif updated to include Gamma and inverse.gaussian families

svydiags 0.6

• Intercept-excluded version of VIF added for all models covered: binomial, gaussian, poisson, quasibinomial, and quasipoisson. The intercept-excluded version is analogous to the one found in other statistical packages for linear models fitted with non-survey data.

• Error corrected in calculation of R-squared in intercept-included version of VIF for non-gaussian models (i.e., binomial, poisson, quasibinomial, and quasipoisson). In previous versions SST in the denominator was not corrected for the mean.

svydiags 0.5

• svycollinear: Updated to include an error trap on the mod and svyglm.obj parameters. If mod is a model matrix, then svyglm.obj must be FALSE.

• svycollinear: Check added when mod is not a matrix but is a data frame. If any column is a factor, then an error is issued that the mod must have factors recoded as zeroes and ones using model.matrix.