Replacement of censored observations by a constant: An evaluation