Working through a messy EFA (factor analysis) in SPSS

In this video I work through a very messy exploratory factor analysis in order to arrive at a workable solution in SPSS.

[elliemoore]

Your videos are great James! I’ve been using them to learn EFA in my PhD data. I’m finding that my pattern matrix is incredibly messy. I mean each item being loaded on a different factor messy. I can see you have a messy EFA example, but for me this example would be the dream dataset for me after working through mine. Do you have any advice on how to work through REALLY messy pattern matrix? Thanks!

[aliceyalenga]

This was great and informative. My problem was actually answered from the replies you had given a student 💕👏🏿👏🏿👏🏿

[shabnammayel]

Thank you for the extremely helpful videos James..is it normal to get negative values in the factor correlation matrix?

[eamonnteh]

This saved my thesis, by the way for the removed items done during EFA do I exclude them when I run a linear regression analysis? Thanks again you’re a big help.

[sitifairuz-mbjl]

Hi sir, thank you very much for the video, I would like to ask, do we really need to ensure the number of factors derived exactly same as the theories? in my case, the first run shows 9 components as opposed to 8 components if based on theoretical framework. however, the component 8 and 9 only consist of 2 items. after several deletions, I managed to get 7 factors with one factor combining 2 separate factors. Should I just proceed with the seven factors and rename the combined factor? I am doing on metacognition and based on literatures, the number of factors are varied across studies. Thank you.

[kirangupta]

Thank you Sir for this video. May I know how to take care of negative loadings in the pattern matrix? What should I do if two factors get loaded on the same component?

[keevanstatz]

Hi Dr. Gaskin -- thank you so much for this really helpful video! I just collected my scale development dissertation data, and unfortunately, it yields a super messy EFA. The theory suggests three factors, but even when extracting specifically for three factors, the loadings are low and there are significant validity concerns (some convergent, but especially discriminant). Therefore, I'd like to conduct an EFA extracting specifically for one factor -- a unidimensional model. Do you have any suggestions as to how to do this on SPSS for the EFA? I've looked around but have struggled to find any information as to actually running a one-dimensional model with its interpretation on SPSS. Thanks in advance!

[itas]

Hi Mr. Gaskin, thanks for this very helpful video. But I want to ask, if there are negative loadings in the Pattern Matrix then I have to reserve the code, right? Then, after doing the reserve code, do I have to repeat from the screening data to EFA or directly to EFA? Please help me. Thank you very much.

[dimitradance]

Hello Mr Gaskin..thank you for your great videos. I would like to ask you something. I did a factor analysis of the Secondary Traumatic Stress Scale and I think that I have to delete one item of the questionnaire in order to achieve a clean factor analysis. I won't use this specific item at the next analyses and results, right?

[tahseenbinte]

Hi,
Thanks for a very informative video.
May I ask: is it fit for sample size of 29?
I have 19 variables, each belong in 3 separate groups such as learner motivation (6), learner experience (6) and learner engagement (7)

[TheFarooqian]

Hello James, i am conducting a research on Industrial Boiler performance and using SEM to validate the theory. I have carried out an EFA USING PCA and would want to do CFA in AMOS. is it necessary that i use Maximum Likelihood technique in EFA to be used for CFA model in amos? Because i am unable to run the EFA using ML. Will appreciate yoir help

[pallavisingh]

Many thanks fro this very useful video. I have done PCA on a sample of 530. In the pattern matrix, 8 variables are very nicely loading on one factor but all of them have negative loadings. What does this mean.

[narsissarafraz]

Dear Dr, James, thanks for great video. I have question. after my EFA, SPSS extracted 12 construct, but it should be only 4.Ihave so many cross-loading and so many high correlation, but even removing them it doesnt work.in one video, you mentioned to remove less than 0.5 item, in another you mentioned to remove those who have less than 0.2 difference, in some you delete high cross loading.high based on what? the difference or highest number? which way is better and get better result to solve the problem? really appreciate your kind guidance

[garimaarora]

Sir i ran an EFA with a sample size of 386 and got 5 factors with decent loadings.
However the total variance extracted came with a footnote that factors are correlated and the squared variances can't be added.
Is there a problem with the factor structure?

[daanishjeelanigaani]

Hi sir,
I am working on a questionnaier which is not completely based on a particular scale and have some multiple answer questions, is it possible to do EFA on that questionnaier.

[ninailia]

Thank you for the video! I have a question. In my research, I use a scale with 14 items to assess the level of receptiveness of the participants to 4 assessment scenarios, meaning that I use the same scale 4 times (and by theory the scale consists of 4 sub-scales). How can I do the EFA? I did a PAF with varimax rotation and 4 factors per assessment scenario, but should I do a PAF for all the 56 items together? Thank you in advance.

[GizzarnT]

Hi, Thank for your useful video. My data got around .6 on Cronbach’s alpha. Also, I have a problem with EFA. The factor should be 7 but it come out only 5. And about Pattern Matrix, the two measurements that should be in separate factor are in the same factor. Also, in one measurement there are 3 var in it. But those three are in separate factor. What should I do? I cannot delete much variables as I have only 3-4 per each measurement. Also in communality, there are around 3 that got extraction lower than .3. After trying to delete it, the Pattern Matrix still in mess. Due to very mess Pattern Matrix, I cannot use CFA as well, I got the message ‘covariance matrix is not positive definite’, but as I see the covariance value it’s not negative value the max value is 1.826 and min value is .189. I tried to change regression value but it’s no use. It disappeared after I deleted one covariant path with very high value. Should I do that?

[shubhamagrawal]

My factor loading is not coming correctly. It is too much scattred and I am using varimax. How can i make it correct?

[toseefmohammad]

Respected sir, what should we in CFA when we got 1. KMO good 2. Eigen value is good with expected factors 3. No issue of parallel loadings. But items of one factor move to the other factor in component matrix.

[inesa]

Thank you for the video but I have a doubt though.. can we not run an EFA with nominal data?