enormous path coefficients etc.

[christian_b]

hey everybody,

i have a problem with interpreting respectively believing in my path coefficients, because they are really enormous.

for my diploma thesis i created a model with 9 latent variables influenced by 2 to 7 formative indicators in each case. these latent variables all influence my endogenous constructs in a formative way.

to test my model i created a sample by myself (i copied a set of 15 cases to get a sample of 100 cases). under a sample of 100 i got the error: "a singular matrix occured during the estimation...". what does this exactly mean for my sample?

after running the pls algorithm i got outer weights between -2.8 and 1.4 for my indicators. the path coefficients of my constructs influencing my endogenous constructs are between -182245529254 and 29680492853!?

another problem is my R², because it´s 1,3 (R²>1?)...

since i´m not the pls specialist by now i hope anyboby can give me some feedback. maybe its my copied sample??

thanks,
Christian

[christian_b]

after further tests and searches it seems to be a problem of multicollinearity produce by my copy-paste sample. real samples from other studies don´t cause this problems. if anybody could emphasise or counter this fact it would be nice.

Christian

[fiasmar]

Dear all,

after i have done FIMIX-PLS analysis, i got the value of R-square greater than 1, eg 1.234. is it true ??
i think range of R-square is 0<R2<1.
i get the value of R-square from the value of FIMIX VARIANCE, is FIMIX VARIANCE same with R-square ??