i just wanted to know if i could use factor scores and crosstab it with demographics (i.e. gender, age, etc.)? I have 69 likert-scale variables and run factor analysis on SPSS. It gave me 10 new variables (types of personality. i just wanted to see the demographics of each new variables. Thanks!!!
3
-
$\begingroup$ Of course you can :) Are you asking how to do it in the software, or whether you can draw a particular conclusion, etc? $\endgroup$– JMSCommented May 11, 2011 at 15:40
-
$\begingroup$ Can you please explain both? :) I saw this study where they use FA to group people's personalities. After they derive the groups, it seems they use Multiple Response Analysis and crosstab it with gender (males and females). As if they change the factor scores into zeros and ones (just a hunch). Thanks in advance! $\endgroup$– jomsCommented May 12, 2011 at 6:15
-
$\begingroup$ I think @rolando2 has you covered on the software front. I'm not familiar with multiple response analysis, but certainly it's reasonable to look at the distribution of factor scores overall, across different groups, etc, however you do it. $\endgroup$– JMSCommented May 12, 2011 at 14:21
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
In SPSS, choose Analyze...Descriptive Statistics...Explore. The factor scores will be your dependents, and a demographic grouping will be your "factor" in this procedure. I'd choose Plots Only to start with and request boxplots. You can get either factor levels together or dependents together: you can experiment.
Later, when you want means and standard deviations and such, I'd use the Summarize command, located in the menus at Analyze...Reports...Case Summaries. You can also experiment with "means [var list] by [var list]/stat anova."
However, I'd caution you that some of your 10 factors are likely to be marginal ones that have small eigenvalues and explain little of the total item variance. Maybe consider making your criteria stricter for factor extraction, such as using a subcommand like "/criteria mineigen (1.5)" or 2.0 instead of 1.0 which is so commonly used. It'll also be a plus to reduce your number of factors when it comes to displaying and explaining your results.
-
$\begingroup$ Thank you so much @rolando2! Great help! I have another question, i tried the descriptive statistics you taught me (factor scores DV and gender for "factor"), but the output is VERY long and i couldn't explain it. Sorry for being a bit stupid here. I was looking for a breakdown in gender (i.e 30% men Factor 1; 80% men in Factor 2; 45% men in Factor 3; etc.). Many thanks in advance!!! $\endgroup$– jomsCommented May 16, 2011 at 7:01
-
$\begingroup$ @joms There is no such thing as being "in" Factor 1: each person has a score on Factor 1. Perhaps the men have scores shifted lower compared to women's--that's the sort of thing you could show with the Explore command. See if you can generate a single boxplot at the end of your output that shows all the Factors split by gender. $\endgroup$– rolando2Commented May 17, 2011 at 0:40