questionnaire independently and return it directly to the researcher in the pre-postagepaid
returned envelope. Subjects are informed that their responses would be
anonymous. Several potential response biases in the questionnaire are checked. The
questions in a survey instrument may affect responses. To protect possible response bias
problems between respondents and non-respondents, a series of t-test comparison of the
demographics between early and late respondents are tested corresponding with
split-in-half technique used to test for non-response bias by Armstrong and Overton
(1977). No statistically significant difference between early and late respondents
demonstrates non-response bias between respondents and non-respondents. Also, the
late respondents represent non-respondents (Nwachukwv and others, 1997).
In this research, all 421 received questionnaires are split into two equal groups
and investigated by t-test. The early respondents are the first group and the late
respondents are the second. Then, 210 responses from the first group mailing are used