Finally, as the objective of PLS is to maximize variance
explained rather than fit, prediction-orientated measures,
such as R-square (R2), were used to evaluate the PLS models
(Chin, 1998). The R2 for the main effects and interaction
models are presented in Table 5. To see how much predictive
value the interaction terms add to the model, the R2
from the main model without interactions (Stage 1) was
compared with the R2 from the interaction effects model
(Stage 2). The additive explanatory power of the interaction
model was determined by calculating Cohen’s f2 effect
size measure (Chin et al., 2003; Cohen, 1988). The results
show that the interaction constructs have a total effect
size f of 0.23 in exploitative innovation projects and 0.07
in exploratory innovation projects, which indicates that
the inclusion of the interaction terms does improve the
explanatory power of the model