Hello everyone,
Whats up??
So, I was doing some tests with genetic (PCA scores), environmental (PCA scores), past climatic variation (PCA scores), and geographical (latitude and longitude) data using SEM analysis and some weird results came across. The determinantion coeficients (Std.all) were higher than 1.0 (100%) and p values very significant (0.000). I looked for explanations in lavaan package and other SEM documentation and videos but I didn't find nothing.
Did someone here already used SEM and get this kind of problem?
Model created and analysis summary are attached
To me it looks like there may be suppressor effects in the model portion for genPC1. Are some of the predictor variables for genPC1 strongly correlated with one another?
The Std.all coefficients are standardized regression coefficients (not: coefficients of determination). The coefficients of determination are given under R-Square. The R-square values are all within the appropriate range between 0 and 1; no improper value there. The standardized regression coefficients (Std.all) can be > |1|. This can happen when predictors are strongly correlated and/or when there are suppressor effects in the model.
Hello Christian, thank you so much for this answer. You helped me so much!!! About your question, yes. I noticed in some tests that Latitude has colinearity with other variables and when I took it out of the model, it works better.
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers