Hello,

I don't know how exactly the imputation feature in SPSS works but it should not increase your sample size but create a set of distinct samples of equal size which serve as data input for the analysis. That is, the analysis is conducted within each of the samples and then the results are compined by the so called "Rubin-rule".

Check this.

All the best,

Holger

@Holger Steinmetz

Thank you for your response. I thought so too. Whatever is done to the dataset shouldn’t increase the sample size. I must have done sth wrong. One of the analyses I ran after imputation showed a number of cases greater than my original sample.

Hello Phuong,

What happened is that the process was repeated 5 times (not an uncommon number of cycles in MI). As Holger correctly noted, the usual method is to analyze each cycle's data set separately and combine the results.

What you're seeing is the agglomeration of the five cycles of data set with imputed values, successively appended (cycle 1 cases, followed by cycle 2 cases, etc.) into your data worksheet file.

Here are three options:

1. Run five factorings, each using a different cycle data set, and attempt to merge the results.

2. Use the (5 times larger) agglomerated data set, as that is likely to give more robust estimates of the correlations among variables for which imputed cases had to be generated. The fact that the sample size isn't the same as your original won't matter except for certain goodness-of-fit tests (such as the chi-square test that maximum likelihood extraction options often report in FA software), nor will the "5 times" size impact estimates of correlations among variable pairs for which no imputation was required.

3. Use bootstrap method from the full agglomerated data set to estimate the correlations for variable pairs involving any imputed values, then use that set of correlations in your factor analysis.

Good luck with your work.

interested

Yes, that is normal: SPSS creates 5 posible values with the imputed data. Then, 5 datasets are pasted long one after another (in a single file), as well as the original one.

You can use that file to run some analyses, and SPSS will give you estimates for the original data (with missing) and each of the imputations, and a final “pooled”

Yes, that is normal: SPSS creates 5 posible values with the imputed data. Then, 5 datasets are pasted long one after another (in a single file), as well as the original one.

You can use that file to run some analyses, and SPSS will give you estimates for the original data (with missing) and each of the imputations, and a final “pooled” estomate. Im not aure if you can run factorial like this, try.

Yes, that is normal: SPSS creates 5 posible values with the imputed data. Then, 5 datasets are pasted long one after another (in a single file), as well as the original one.

You can use that file to run some analyses, and SPSS will give you estimates for the original data (with missing) and each of the imputations, and a final “pooled” estomate. Im not aure if you can run factorial like this, try.