ered satisfactory (see the discussions in Devellis,
1991; and Pedhazur and Schmelkin, 1991). These
resulting scales (SHOULDBE and HOWITIS)
were used as appropriate dependent variables for
analysis with all items weighted equally. The
decision to give items equal weights in scales has
extensive support in the literature and is consis-
tent with the recommendations of Babbie (1992),
Comrey (1988), Nunnally, (1978), and Pedhazur
and Schmelkin (1991). Specific measures used to
construct the two dependent variables are given
in Table I.
Results
Of the 691
students involved in the study, 379
were male and 294 were female. Eighteen
subjects did not report their sex. The average age
of freshmen in this sample was 18 while the
average age of seniors was 21. Eighty-five percent
of the students reported that they had taken at
least one course in which a portion of the time
had been allocated to discussing ethics. Twenty-
four percent of the students had taken a formal
course in business ethics. Because of missing
values, responses from 33 of the subjects were
eliminated from the sample. Therefore, with the
exceptions
of the correlational data, the following
results are based on the responses of 658 subjects.
In doing the analysis the General Linear Models
(GLM) procedure of SAS was used. The GLM
procedure is highly versatile, allows for many
different types of analysis, and can handle both
classification and continuous variables.
Correlation coefficients describing the rela-
tionship between variables are given in Table II.
Significance of each correlation is given in
parentheses underneath. Because of missing
values the
N for different correlations varies
slightly. The means
and standard deviations of
the dependent variables are given also. Note that
the lower numbers indicate that respondents
tend to strongly agree that there should be
(SHOULDBE) or there is (HOWITIS) a strong
relationship between good business ethics and
positive business outcomes.
Because the two dependent variables are sig-
nificantly correlated, a multivariate analysis was
conducted to determine the overall effect of SEX
(male or female), CLASS (freshman or senior),
and ETHICPRT (exposed to ethics in some
portion of a course or never exposed to ethics
in a course) simultaneously
on SHOULDBE and
HOWITIS. This allows a simultaneous test of
differences among multiple dependent variables.
Many researchers recommend that MANOVA be
conducted first to determine whether there are
overall significant differences among groups
before proceeding with more specific univariate
analysis (Pedhazur, 1982). Although there are
many tests available for this purpose the most
widely used is Bartlett’s (1947) test of Wilks’
Lambda.
The hypothesis of no overall effect of SEX was
rejected (Wilks’ Lambda = 0.9843389,
F
2, 649
=
5.1629,
p < 0.006). The hypothesis of no overall
ACADEMIC-CLASS effect was also rejected
(Wilks’ Lambda = 0.964741,
F
2, 649
= 11.8594,
Do'stlaringiz bilan baham: