Dear forum-users,
using smartPLS for the first time, a couple of problems emerged. I hope that you can help me and I’m thankful for every answer (the analyzed model is third-order with reflective indicators and consists of seven dimensions of social responsibility – three of them build up by two or more factors and four of them by single factors; all in all there are 34 indicators and Wolds hierarchical design as described in Wold (1982) is utilized; N is 305). In order to evaluate the model, I tried to use the quality-criteria as described in Ringle (2004).
My questions concerning the quality-criteria for the structural model are:
1) The listed values for dimensions R-squares are extremely high (all above 0,999); all pathcoefficients are positive (as they are supposed to be) and six from seven are above 0,1 (as Lohmöller sugested). Inner models t-statistics are high enough, all of them being significant (Bootstrapping: cases: 305 (equal N), samples: 500, one-sided t-test (alpha: 5%)).
LV correlation inside dimensions (up to 0,88 in one of them) is higher than outside dimensions (most of them from 0,1 to 0,35, two correlations are as big as 0,55 and 0,60). Some of the correlations are slightly negative.
Controlling the impact of the exogenous LVs on the endogenous LVs ((R square included-R square excluded)/1-R square included), results differ from slightly negative up to 3,2.
Is there an explanation for the extremely high R squares and the impacts? First of all I suspected multicollinearity but after checking the signs and the significances, I’m quite unsure. For both possibilities (multicollinearity or something else): which literature can I citate and which consequences are there for the following interpretation of the quality-criteria?
2) All Q squares are positive ( between 0,33 and 0,63) but some of the q squares ((Q square included – Q square exluded)/1 – Q square included) are negative (-0,08; -0,06; -0,01; -0,09; 0,007; -0,06; -0,01; 0,005; 0,32; 0,51; 0,01) - how should I interprete these results? Is there a detailed description for interpretation in Chin (1998)? I ordered his articel but still have to wait for its deliverance.
My questions concerning the quality-criteria for the outer model is:
3) Using bootstrapping for significance-tests: one-sided tests are sufficient if there is an a-priori sign determination – or did I get it wrong?
Thanks for your answers!
Best regards
Marvin