[#5175] predict() method for sampleSelection

What is the best user interface?

In order to be consistent with predict.glm() and coef.selection(), I suggest:

predict( object, part = "outcome", newdata = NULL,
type = ifelse( part == "outcome", "unconditional", "link" ) )

If argument "part" is equal to "selection", the predicted selection variable is returned and argument "type" can be "link" (meaning that E[w*|Z] = Z %*% gamma is returned) or "response" (meaning that E[w|Z] = pnorm( X %*% beta ) is returned).

If argument "part" is equal to "outcome", the predicted outcome variable is returned and argument "type" can be "unconditional" (meaning that E[y|X] = X %*% beta is returned) or "conditional" (meaning that E[y|X,Z,w=1] = X %*% beta + dnorm( Z %*% gamma ) / pnorm( Z %*% gamma ) * lambda is returned).