The pattern of correlations and reliabilities was nearly identical
between Samples 1 and 2, and thus summaries of results are presented
here, and statistics are reported in Table 2. We first performed
confirmatory factor analyses on each facet scale (PE and
WE) separately. Similar to the pilot study, for both PE and WE, the
hypothesized four-factor model showed better fit than a singlefactor
model in both samples (Table 3). However, the four subscales
within each model were correlated strongly (factor
r ¼ .80e.93), and the single-factor model for each facet showed
adequate fit. To facilitate interpretation, we chose to aggregate
across subscales to form an overall score for each facet. Reliabilities
for WE (a ¼ .98) and PE (a ¼ .97) were high.
However, there was also a high correlation between the PE
and WE scales overall. Thus, we tested a model with a single
general factor for all of the items (PE and WE together) as well as
a model with two general facet factors (i.e. PE and WE). The twofacet-factor
model fit slightly, but significantly better, than the
single-factor model; however, neither fit well (Table 3; though
they are not directly comparable to the other models because
they are not nested). Further, because we hypothesized that
equality (or inequality) of values for productivity and wellbeing
can impact work-related pain, we decided that for conceptual
reasons, it was best to keep PE and WE facets of climate separate.
RSM can be used even when predictors are substantially correlated,
because RSM is concerned with not only correlation but
agreement. A high correlation between variables does not
necessarily mean that the values assigned to each variable are
the same. For example, ratings of “5-4-3” on one variable and