I have developed several question on Likert scale, but each question is mutually exclusive, but of course they intend to measure the same research question.
Hello Rizwana,
do they measure the same latent variable? Do you have a theory how many factors there are?
Hi Timo,Thank you very much for your response, Yes they measure the same latent variable,actually there are two research question one is measuring the Experiences it has 8 factors, and other question intend to find the hidden reasons of women low progress in job Market it has 15 factors.
Do you mean 8 items and 15 items or factors?
Did you consider using the lavaan-package in r?
http://lavaan.ugent.be/
Yes 8 Items, I realized that mistake :), I did not know about Lavaan :(
Hey,
here is a very good introduction from the book "Mastering Scientific Computing with R"
https://www.packtpub.com/books/content/structural-equation-modeling-and-confirmatory-factor-analysis/?utm_source=ArchitC&utm_medium=referral&utm_campaign=article#more
This issue has come up so often here that I have collected a set of resources on it:
https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items
Nicholas Peiper, Matt Jans and David L Morgan Thank you much for your feedback and concern. Your effort is much appreciated.
Hi Rizwana - Yes indeed, you can use factor analysis. Your questionnaire has 8 + 15 items = 23 items total . . . So you could use exploratory factor analysis to determine ideal groupings from the items in your questionnaire. You could also use forced factor analysis to see how the items collapse into 2 distinct groupings. When using factor analysis, the recommendation is to collect at least 2.5 times as many completed questionnaires as you have items. So with a total of 23 items you would need about 58 completed questionnaires for the purposes of factor analysis.
Hi James Hunt, Brilliant, Thank you so much for your detailed answer. Much appreciated :)
James - Sir can you please explain why are we adding 8 + 15
Like i AM using a multidimensional condom attitude scale it has 25 question items so what does that suppose to mean does it mean that i have to 2.5*25 = 63 completed questionnaires for the purpose of factor analysis.
Further what is the difference between forced and exploratory factor analysis?When do you use exploratory and forced factor analysis?
Likert scale is most ideal for Factor analysis. You need to have a set of interrelated items with good facial and construct validity and with a cronbach Alpha value above 0.7.
As far as I know, the issue of how many items one needs to construct a measurement model has no agreed-upon answer. In addition to the suggestion here of 2.5 observation per parameter (coefficient) in a model, I have also seen 5 and 9 as suggested standards.
In general, these estimates are generated using simulation models to detect how stable the solutions are, rather than statistical rules.
I made a mistake when I said "how many items one needs to construct a measurement model." I meant to say: "how many observations one needs to construct a measurement model." Yes, one needs at least 3 variables (items) to do a factor analysis, which is different from the N that one needs in order to get stable estimates.
Also, unless you believe that the concepts underlying your scales are uncorrelated, I would recommend oblique rather than orthogonal (varimax) rotation.
Ashish - if your instrument has 25 items, and you want to conduct factor analysis on the data, you would need an absolute minimum of 63 completed questionnaires (25 X 2.5) but of course, more responses are desirable . . . As suggested above, some recommendations suggest 3 x n as a minimum starting point.
The difference between exploratory and forced factor analysis, is that with exploratory, the 'ideal' number of factor groupings is presented in the factor run, whereas with forced factor analysis, you can force a solution for any number of factors so you might request a solution for 3, 4 or 5 factors for example.
Likert Scale is suitable for FA. Pallant (2005), however, recommends that sample should 150 or more
Thanks Frederick - Yes. In fact the sample size really depends on the number of items in the questionnaire. This effectively dictates the minimum number of completed responses required in order to conduct meaningful Factor Analysis.
To carry out meaningful Factor Analysis 500 cases is very good, 1000 or more is excellent.
Following up on what James said, the key issue is the number of "parameters" that you are going to estimate, which is based on the number of questions you asked. Most people attempt to get so-called simple structure structure, where each item loads on one and only one factor. If that is so, the number of parameters is equal to the number of items (.e., one loading per item), unless you allow for correlated factors. In the latter case, you also have to count a parameter for each correlation among the factors (which in the case of two factors, would be just one correlation).
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS