Dear researchers
I am using spss
is normal distribution is required in exploratory and confirmatory factor analysis in survey?
I have at least 4 items for each axies and 7 at maximum, I ask when we can't do the exploratory and confirmatory factor analysis in survey?
what are requirements? conditions?
If you were running a CFA, you'd need to satisfy both uni- and multi-variate normality. See examples:
Article Validating a New TQM-Benchmarking Measurement Model in an In...
Article A Confirmatory Factor Analysis for SERVPERF Instrument based...
Thanks,
Ali
Hi Abdelkarim,
as an addition to Ali's post, there are some approaches how to easily correct for non-normality, e.g., the Satorra-Bentler or Yuan-Bentler correction. Most software packages allow to easy apply there correction.
Best,
Holger
Reliability analysis is conducted to check the homogeneity between variables. In exploratory factor analysis (EFA), multivariate normality is not required.
Yes you need to satisfy the condition of normality. Your data must be normally distributed to obtain valid parameters. Similarly, other requirement of EFA and CFA is the sample size, some rule of thumb provided that, 100 samples to be minimum while others requires upto 300 samples
Hi Abdelkarim
I tend to side with Holger on this issue. We now can easily correct for the deleterious effects of non-normality and/or crude measurement by using maximum likelihood estimation, or we can calculate the correlation that we would expect if our variables were normally distributed by using polychoric or tetrachoric correlations.
Of course this does not mean that we shouldn't care about the distribution of scores, rather that deviations from normality are not the threat that they once were to the validity of our analyses.
Mark
Thanks Mark,
an additional point which just came to my mind is that of course one has to check if the variable of interest is normal "in principle" but deviates from the distributional form or whether it's essentially non-normal (e.g., poisson, or gamma). For instance, often people are not aware that count variables are seriously non-normally distribution--even in theory and require other methods.
Best,
Holger
Yes your data should be normally distributed. In addition to that check crobach alpha and KMO to check the validity of data for factor analysis. After applying factor analysis, the communalities should be greater than 0.3.
As long as the population is normally distributed, the sample as well must satisfy the normality condition in addition to other requirements for Factor Analysis, e.g. Sample size as mentioned by @Ado ABDU Bichi.
Here you have some recommendations to conduct EFA/CFA with survey data:
1) All your variables are continuous AND multivariate normal distributed: In this case you can use the standard Maximum Likelihood estimator.
2) All your variables are continuous BUT NOT multivariate normal distributed: In this case you have to use Maximum Likelihood with a "correction for non-normality" (e.g., Satorra-Bentler).
3) Your data contains a mixed of continuous AND ordinal variables: In this case you have to use Diagonal Weighted Least Squares as the estimator. The most well known estimators in this category are WLSMV and ULSMV.
4) All your variables are ordinal: Same as 3).
Greetings,
Roberto
Hello all,
I am a bit confused due to different comments.
Well, I have a single factor containing 11 items (Likert rating). For the EFA, I am using SPSS (maximum likelihood) and I use lavaan and Amos for the CFA. I've got three questions:
1. KMO and Bartlett's tests' criteria are met while the normality tests (Kolmogorov-Smirnov and Shapiro-Wilk test) are not met (they are both significant). So, am I good to keep up the EFA or shall I need to use Satorra-Bentler or Yuan-Bentler adjustments (if yes, what software do I need to use)?
2. Should I be checking the normality for each item or the variable?
3. For the divergent validity, I use two other variables aside from my main questionnaire. Do they also need to be distributed normally as well?
Thanks for your time,
Sara
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis