ability to make a more complete evaluation of differences between groups. Fifth, two
indices had high levels of non-response and as a result had to either be withdrawn from
the study (normative beliefs) or reduced to a one-item question (subjective norm). Non-
response may be indicative of sampling bias. This can be due to non-random error where
certain types of people are more likely to not respond than others. I do not know how
these cases differ than those who did respond. Lastly, pilot test participants were all part
of one user group. The instruments would more reliably measure the constructs under
study if I pilot tested the instruments with cases representing all user groups.
Data analysis
I tested for differences between the three groups based on the outcome variable
TAP using a Mann-Whitney U test. I determined if the groups were different based on
the predicator variables to measure the first, second and third hypotheses. I used the
Mann-Whitney U test to test difference between groups with ratio, interval and ordinal
data. I did not use a student t-test to test for differences between groups with interval and
ratio data because the groups had small samples, the sample size as a whole was small,
and the number of smalls in the groups were uneven. The Mann-Whitney U test is a more
conservative measure of group differences than the student t-tests (Sheskin, 2004). I used
Fisher's exact test to test difference between groups with nominal data. I ran a logistic
regression to test the fourth hypothesis. I conducted this type of analysis because the
outcome variable was converted from an interval measurement to a nominal level of
measurement.
Definition
* Attitude: a person's judgment that performing the behavior is good or bad
* Behavior: Behavior is the manifest, observable response in a given situation