Should providers of treatment be regarded as a random factor? If it ain't broke, don't "fix" it: a comment on Siemer and Joormann (2003).