Results
Confirmatory factor analysis of the clinical
leadership survey
A series of three measurement models were evaluated in
a cross validation process of the CLS. The sample
(N = 480) was randomly divided into two samples for
this purpose. A confirmatory factor analysis of the original
CLS was conducted with the first sample
(n = 244), using an iterative process to validate the
factor structure of the newly developed measure. This
analysis suggested the need to revise the measure to
obtain acceptable fit indices. A measurement model
consisting of 15 items (three items per subscale)
achieved a very good fit in the original sample and was
supported in the second sample in the cross-validation
procedure. The iterative process involved examining the
error terms, the magnitude of the correlations between
the items and their factors, and reviewing the correlation
matrix for high correlations among items. The final
15-item measurement model was theoretically consistent
with our conceptual model and was evaluated
using maximum likelihood estimation. Standardized
loadings between items and their associated factors
ranged from 0.48 to 0.86 (Table 5). The loadings between
each first order factor and the overall CL latent
variable ranged from 0.55 to 0.98 (Figure 2). The
magnitude of these paths was strong supporting the
construct validity of the measure.
The final 15-item model resulted in a good fit with the
observed data [v2 = 128.678, d.f. = 85, P = 0.002,
comparative fit index (CFI) (0.96), Tucker–Lewis fit
index (TLI) (0.95) and root mean square error of
approximation measure (RMSEA) (0.05)]. CFI and TLI
coefficients above 0.90 and RMSEA coefficients of 0.05
or below represent an excellent fit (Musil et al. 1998).
Although the chi-squared was significant, this is often
the case with large samples and therefore must be
considered among a set of indicators, each of which
adds to a meaningful interpretation of the results (Hoyle
& Panter 1995). Cross-validation of this measurement
model using sample 2 (n = 236) yielded similar results
(Table 6).sample 2 (n = 236) yielded similar results
(Table 6).
Further support for the concurrent validity was
determined by a significant positive correlation between
the CLS and the global measure of clinical leadership
(r = 0.45). The results of this analysis provided preliminary
evidence for the construct validity of the CLS.
We then proceeded to examine the hypothesized relationships
in our nomological network to obtain further
support for the construct validity of the CLS.