To investigate how the different components relate to the general construct of relationsoriented
leadership behaviour, we distinguished between three plausible scenarios. A series of
confirmatory factor analyses were performed and compared in terms of goodness-of-fit to
determine the most appropriate scenario (Bentler & Bonett, 1980; Kernis & Goldman, 2006).
The first scenario says that although relations-oriented leadership behaviour may consist of
conceptually distinct components, these are not empirically distinguishable due to high
interrelations. In this case, a one-factor solution representing a general relations-oriented
leadership behaviour construct should give the best goodness-of-fit results. The second
scenario represents a situation where relations-oriented leadership behaviour is a 16
multidimensional construct, consisting of five different components. This scenario
corresponds to a five-factor solution. Finally, in the third scenario it is assumed that the five
components are conceptually distinct, but interrelated facets that are influenced by a general
relations-oriented leadership behaviour construct. In factor analytic terminology this equates
to a second-order factor analysis, with the five components being indicators of a latent
relations-oriented leadership behaviour construct. The goodness-of-fit indices for the three
scenarios are shown in Table 2. Based on the suggestions of Hu and Bentler (1998), we focus
on four different goodness-of-fit indices: Bentler’s (1990) comparative fit index (CFI), the
root-mean-square error of approximation (RMSEA), the standardized root-mean-square
residual (SRMR) and the χ²-difference test. The standard χ² test statistic is not taken into
account because of its sensitivity to sample size (Marsh, Balla & McDonald, 1988; Hu &
Bentler, 1998). The one-factor model provides a poor fit to the data according to the CFI
(values close to 0.95 indicate good fit) (Hu & Bentler, 1998) and the RMSEA (values ≤ 0.08
indicate fair fit) (Browne & Cudeck, 1993; Hu & Bentler, 1998). Only the SRMR is below the
recommended value of 0.08 (Hu & Bentler, 1998). However, as compared to the one-factor
model, the five-factor model shows a much better fit to the data (∆χ² = 1452.00, ∆df = 47,
p<0.001; and CFI=0.94, RMSEA = 0.074, SRMR = 0.038). The second-order factor model
shows an equally good fit as compared to the five-factor model (∆χ² = 2.36, ∆df = 5, not
significant; and other goodness-of-fit indices almost identical to the ones of the five-factor
model). This indicates that the more parsimonious second-order model should be preferred
and supports the conception of relations-oriented leadership behaviour as a multidimensional
construct, of which the components are related due to the existence of a higher order construct
(Kernis & Goldman, 2006). The factor loadings for the second-order factor model are shown
in Figure 2. They are all statistically significant (p<0.001) and higher than 0.40.
การแปล กรุณารอสักครู่..
