How to handle extreme multivariate skewness and kurtosis in Confirmatory Factor Analysis?

CFA
1

I’m working on a Confirmatory Factor Analysis (CFA) for a 25-item scale. I want to test a two-dimension model. But when I checked the data, I found really high skewness and kurtosis values. I’m worried this might mess up my CFA results.

Here’s what I did:

data_subset <- data[, c('item_1', 'item_2', ..., 'item_25')]
data_subset[data_subset == 999] <- NA
clean_data <- na.omit(data_subset)

skew_result <- calculate_skewness(clean_data)
kurtosis_result <- calculate_kurtosis(clean_data)

The output shows:

Skewness: 1710
Kurtosis: 3668

These numbers seem really high. Is this normal? Should I transform my data somehow? Or is there another way to deal with this in CFA? Any advice would be great!

2

wow those numbers are crazy high! you might wanna double check your calculations first. if they’re right, consider robust methods like satorra-bentler or bootstrapping. also, look for outliers or data entry errors. what kind of scale is this? some scales naturally have extreme responses. don’t give up, there’s usually a way to handle it!

3

Whoa, those skewness and kurtosis values are off the charts! :open_mouth: Have you double-checked your calculations? Those numbers seem unusually high, even for severely skewed data.

If they’re correct, you might be dealing with some serious outliers or data entry errors. Maybe take a closer look at your raw data? Sometimes a few extreme values can really throw things off.

For CFA, non-normality can definitely be a headache. Have you considered trying robust estimation methods? Something like Satorra-Bentler might help. Or you could look into bootstrapping - that can sometimes work wonders with funky distributions.

Just curious, what kind of scale are you working with? Is it something where extreme responses are expected, or is this totally out of left field?

Whatever you decide, don’t let those numbers discourage you! CFAs can be tricky, but there’s usually a way to work through it. Keep us posted on what you find out!