Factor analysis on likert scale questionnaire Help Needed!!!

#1
Hi,
I am currently at the process of finishing my thesis and one of the means of gathering data I have used was a questionnaire with a likert scale which asked repondents to rate different factors as BAD, POOR, WELL and EXCELLENT
Now after doing analysis in a very simple way comparing means my supervisor told me that I cannot do it that way and I have to do factor analysis. After reading today all day about it, I feel completely lost and I am not sure in which way the data should be prepared and how what would output mean. Using trial version of statistiXL.

So far the data is organised as Bad-2 Poor - 0 Well -12 Excellent - 4 which is amount of who ticked that option. That does not deliver correct mean. unless multipling values as BAD (1) WELL (2) etc.. I can get mean that way, but not sure how to get excell to get correct output.

Could you please suggest anythingI would really aprecciate any help on this matter.

Regards

Nikolay
 
#2
I also believe that 16 subjects for about 50-60 variable items is too poor. Am I correct? is there any other way to the analysis of the data I have got. I was suggested to combine a set of questions by theme and use the average for them, but I cannot find the description of such technique anywhere on internet could you point me at it? or provide me its name?

Thanks

Nikolay
 
#3
You're right. You have too few subjects for a Factor Analysis. You need a minimum of 5 subjects per variable.

If you need a really good book about Factor Analysis, I recommend Larry Hatcher's book. I think he talks about creating 'Factor Based Scores,' which is what you're talking about in averaging variables.

Other than that, it is hard to say what your other analysis options are without knowing the data and design issues.

Karen
 
#4
Hi,
Thanks for the reply, I am doing research on project management methodologies.

All my questions are asking respondents to rate how does methodology X handle____?

Clearly defined requirements (bad, poor, well, excellent)
Poorly defined requirements (bad, poor, well, excellent)
High level of customer participation (bad, poor, well, excellent)
Low level of customer participation (bad, poor, well, excellent)

What can I do with this data? what techniques can I use to analyze it? I used means for each but it is incorrect approach as I have learned... I was suggested to group questions and use average score for these groups, but I don't understand how to do it and I cannot find description of it on internet (dunno the name of technique).

Would mean, SD, median and interquartile range be sufficient? or do I need to do some kind of test and correlations. (though not sure how I would interpret such kind of data)

Thanks

Nikolay
 
#5
Whether descriptives like mean, sd, etc. are sufficient depends on your research questions. What are you trying to find out? Which mean comparisons did you originally do?

Karen