PSY 370 Lecture Notes - Lecture 23: Correlation And Dependence, Factor Analysis, Psy
Document Summary
Sometimes the very best model we can come up with is still not a good fit to the data. We specify the model (a set of equations) and propose that it will describe our data pretty well. We then compare the correlation matrix predicted by our model to the real correlation matrix. Goodness-of-fit indices tell us how far off the real data is from our model. How far off are the model"s predictions on average. Cfi > . 90 (ranges from 0 to 1, and we want this to be big) Rmsea < . 10 (this analyzes the error, so we want this to be a small number) You will also get a chi-square value, which isn"t all that meaningful by itself, but . "all models are wrong, but some models are useful" (or better than others) Just like we test alternative explanations in experiments, we test alternative models in factor analysis.
