Response bias and the need for extensive mail questionnaire follow-ups among selected recreation samples.