captured by the construct in relation to the amount of variance due to
measurement error” (Fornell and Larcker 1981: 46).
The formula used for calculating AVE is:
sum of squared standardised factor loadings
__________________________________________________________
sum of squared standardized factor loadings + sum of error variances
(Fornell and Larcker 1981: 47)The formula used for calculating construct reliability is:
squared sum of standardised factor loadings
____________________________________________________
squared sum of standardised factor loadings + sum of error variances
(Fornell and Larcker 1981: 46)Discriminant validity
Discriminant validity concerns the extent to which individual latent factors
measure discrete constructs, that is, the extent to which the individual factors are
truly distinct. Kline (2005: 60, 73) points to correlations between latent factors in
the region of >0.85 and >0.90 as being indicative of constructs which are not
sufficiently distinct from one and other. Fornell and Larcker (1981: 47) provide
the rule of thumb that to ”fully satisfy the requirements for discriminant validity”,
then the estimate for the proportion of AVE (average variance extracted) for each
individual latent factor should exceed the proportion of the estimate for the
squared correlation between the factors.
Nomological validity
Following Hair et al. (2006: 811), in order to assess the adequacy of the
nomological validity for the measurement model, the relationships between the
constructs were examined to ensure that the measured correlations between
latent factors are not counter to what is implied by the theory behind the model.
In particular, the correlation estimates between the latent factors are checked for
statistical significance. Elsewhere, Shumaker and Lomax (2004: 106) consider