I'm doing an analysis on a set of 90 test results in 7 different tasks (90 students were given a set of 7 tasks). The results were not given in scores but rather in grades of four levels (A, C, E and F input as 20, 15, 10 and 0). The tasks were designed to represent a certain skill used in different fields (fields that the students had studied; engineering, economics and so on) and one of the aspects we want to explore is how the differences and similarities of the results between fields could be explained by some kind of latent variables.
We've done a factor analysis with MLE for one, two and three factors and a principal factor analysis for three factors that yielded a similar result as the the factor analysis with MLE for three factors.
But now I've tried to find a variant of the method but for ordinal variables and found this: http://www.ats.ucla.edu/stat/stata/f...ategorical.htm but it says that "if both variables take on more than 10 values a Pearson's correlation is calculated." which all the variables that I have do so is there a point of using another method for an analysis that includes ordinal variables?
Last edited by Danrob; 01-02-2017 at 05:43 AM.
Advertise on Talk Stats