How to obtained composite index using factor analysis on likert scale data I am analyzing the influence of social capital on household income. I have collected data through a questionnaire survey for my study.Different dimensions (variables) of social capital were measured using 5-point multiple likert items.The scores on each item was summed up to obtain the single value for each variable.Now i have 6 such variables and I want to perform factor analysis on these 6 variables to obtain a single social capital index for each household.My question is that can I summed up the factor scores (obtained through factor analysis) to obtain social capital index? if my approach is not appropriate, please suggest me some alternatives. I have to use this social capital index in multiple regression for further analysis.
 A: There are 2 things that you are trying to achieve:
1. Summarise each of the 6 variables from their individual items
2. Summarise the 6 variables
For 1, you used the sum of the items under each variable as a representative score. This provides an uneven basis to compare the scores across the variables. A better way would be to use the mean of the items belonging to each of the 6 variables as the representative score for that variable. Also, have tou tried running a factor analysis on the items itself to determine if your 6 factor model is a good fit for your data?
For 2, I am unclear as to how you intend to use factor analysis to create a social capital index. Factor analysis would outline the relationships between latent factors and the variables. Here you would be assuming that they all belong to 1 latent variable which might or might not be the most appropriate factor model for the 6 variables.
For summation of the factor scores, you are assuming equidistant between the likert scales and unidimensionality of all the 6 variables which might be a stretch. An alternative would be to use the 6 factor scores in your multiple regression.
