The posttest samples for this first hypothesis test contained
six year average monthly quality performance results from the
four companies. Eight groups of equal sample size (ni=8=72) were
collected, containing the monthly average product quality score
and the monthly average quality issue rate results for each of the
four companies. To ensure that the characteristic of the samples
matched the characteristic of the population, stratified random
sampling of various product model variations was applied.
Dependent variables were the monthly average product quality
score and monthly average quality issue rate. The independent
variable was the actual usage of a KM/QM strategy (KM/QM
administered, KM/QM not administered). Company B received
the treatment (KM/QM strategy administered) and was included in the two experimental groups of this quasi-experimental
design. Companies A, C, and D never received the treatment
(KM/QM strategy not administered) and were included in the
six control groups of this quasi-experimental design. Using
control groups in an experiment increases the internal validity
of the experimentation (Salkind, 2012). The quasi-experimental
design method was selected since it allowed an evaluation of
the effects of variables associated with the hypotheses after the
fact. Quasi-experimental designs are therefore also referred to as
“post hoc research” (Salkind, 2012). To determine if there was a
difference between the groups before implementing the KM/QM
strategy at company B, a pretest was performed for three years
before the KM/QM strategy was administered and compared
to the other sample groups. Descriptive statistics and a single
factor analysis of variance (ANOVA) was used to test the two
dependent variables, product quality score, and quality issue rate,
to determine whether or not there was a significant difference
between the groups. In case there was a significant difference, an
analysis of means (ANOM) was performed to characterize the
differences. The level of significance used for this study was α=
0.05, meaning there was a 5% chance that a null hypothesis was
rejected when it was actually true, also referred to as a type one
error (Salkind, 2012). Microsoft Excel was used to perform the
ANOVA. Minitab 16 statistical software was used to perform the
ANOM and the residual analysis. ANOM graphs were generated
with Minitab 16 statistical software.