Statistical implications of selectively reported inferential results