Factor Analysis and Principal Component Analysis Using SPSS | A User-Friendly Guide

In this video, I demonstrate how to run Factor Analysis and Principal Component Analysis Using SPSS. I will provide a user-friendly discussion of eigenvalue, Kaiser-Meyer-Olkin (KMO) test and Bartlett's Test of Sphericity, communalities, factor loadings, rotations (orthogonal vs oblique), and the determinant of the correlation matrix.
For more information, please see:
www.routledge.com/Quantitativ...
us.sagepub.com/en-us/nam/stat...
Examining the factor structure and its replicability across multiple listening test forms: Validity evidence for the Michigan English Test: journals.sagepub.com/doi/full...

[@Tahlilgar.amar.1402]

It was great. Thanks for your full explanation ❤

[@niloufar.alizadeh]

Thank you for providing a clear and helpful description.

[@hinadiakbar7292]

Very informative and helpful. Thank you

[@user-hh3ik4kd9c]

Thank you! Very helpful!!

[@user-wh1sv4bx8u]

14:22 Hi sir, some people say that the extraction value must be more than 0.5, if it is below that then you have to eliminate that variable. but you make a big deal out of it

[@sahelmoein4377]

Thank you for your tutorial. I have a question and I really appreciate if you help me with that. if I run a principal axis factoring, and I want to reduce the number of factors using Parallel analysis, should I compare the eigenvalues generated from the random data set (from Parallel analysis, created by R, and it is set on factor , not principal component) with the eigenvalues in the first column or with extraction sum of squared loadings ?

[@ayushmaanchoudhary1900]

Awesome

[@saeedaryadoost7555]

Very nice

[@shaidathuljemin7142]

hye Dr.very brief explanation. i wnt to ask should we reverse code the item, since got 1 item r different from the rest, which is 'Standard deviations excite me'. then, mayb we can reduce the probability of cross-loading items in the pattern matrix?

[@Tushi_07]

My correlation matrix result is not showing. Its just showing " this result is non positive definitive ". What to do now?

[@ismatzahra3]

I loved your video and how you explained each component in the table. However, I am a bit confused about the Principle Component and Principle axis factoring, what is the difference between them and let's say if I am looking into the student's perception of Game-Based Learning, which one should i use from both of them. Also, regarding the rotation, can i use Varimax instead of Promax and vice versa? i would be happy if you could reply to this question. Thank you so much, Dr.

[@user-bc7fq6cb1v]

Hi I didn't get components number are which variable that you imported

[@ahmedel-sayed8744]

Thanks can you pls provides us by the data which used (sav)

[@mnorberta24]

Hello! Thank you so much for the great video. What does it mean if the determinant is smaller than 0.00001? And what should I be doing? Do I remove some items?

[@VahidAryadoust]

The data are not suitable for factor analysis due to multicollinearity. You might want to find the cause for multicollinearity to resolve the issue.

[@zerihunkura4828]

Thank you so much for the the interested video. how can I compute the wealth index from different variables which have different scale of measurements like household's ownership of selected assets, such as Television and sofa; materials used for housing constrictions; and types of water access and sanitation facilities using PCA in spss? or what is your recommendation for me?

[@VahidAryadoust]

You could save the PCA scores in SPSS; or use the Rasch model to compute scale scores based on the different variables you got.

[@ericnovotny7381]

Thanks for the video! I followed everything (using my own data) and got similar output to yours, until the step where you ran Factor Analysis. I got the error "Attempted to extract 3 factors. In iteration 25, the communality of a variable exceeded 1.0. Extraction was terminated." How can I address that?

[@VahidAryadoust]

possible causes: error in your data (e.g., miskeyed-in data); small sample; large number of items. In addition to resolving these, please also increase the number of iterations to 100 as explained in the video.

[@ediblealuminium1835]

Hi Dr., thank you so much for a really useful video. I've followed the steps but I have the results: "This matrix is not positively definite", and "when components are correlated, sums of squared loadings cannot be added to obtain total variance". I'm confused as to what this means, is my data not ok for a PCA?

[@VahidAryadoust]

This happens mainly because of the small sample size and/or a large number of items in the analysis. I'd suggest you extend your sample size, or if it is not possible, validate your instrument using the Rasch model. Here is a video: kzbin.info/www/bejne/gV67mWWhp56Xq7s&

[@thomneijzen]

Hello Sir. I was wondering if I could ask you a question about the determing of factors in the part of principal axis anaylsis (27:13 and on). I was wondering why we should take a look at the 'extraction Sums of Squared loadings' (total) column, for determining the number of factors in the principal axis analysis (instead of looking at initial the eigenvalues such as within the principal components analysis). Thanks in advance.

[@VahidAryadoust]

The two sections in the results provide you with similar / even identical results. You could check out either of them.

[@sushmitadebbarma4980]

Very insightful. Since my attitude scale is multidimensional, factor analysis(EFA) is necessary before determining reliability; unfortunately, factor analysis is not possible with the small sample size of 40 in my pilot survey. In the absence of factor analysis, reliability(cronbach's Alpha) is very low. Should I do factor analysis in the main survey? If so, how should the pilot survey's reliability test be handled? Could you please advise me on what to do next? With appreciation

[@nanaoseibonsu3798]

Do content and face validity. Revise the instrument based on the feedback received from supervisors and friends etc. Go ahead and gather large data and then do factor analysis. You then analyse your main data from there. update me on what you did

[@davidaparicio2313]

AT 29:34 In the pattern matrix box factor 3 has 3 negative numbers (-.877, --.498, -.316) does it mean that we should delete that factor?

[@VahidAryadoust]

No, you'd better look at the content of the items. eg, i'd expect items measuring stress would have negative loading coefficients compared with items measuring happiness. Also, check if they will need to be reverse-coding.

[@ayelakhan2494]

which table do I use to report i the analysis between communalities or component matrix table? im confused about what to do after the kmo and bartletts analysis. Can you please help

[@VahidAryadoust]

Report both of them.

[@ericasare8864]

Very insightful. Dr. can an exploratory factor analysis be done on only seven items to make up a scale?

[@VahidAryadoust]

Eric, yes, EFA will be suitable for that. Ensure each factor will be indicated (measured) by at least 2 or 3 items.

[@ericasare8864]

Thank you for your response

[@talzabidi1569]

Hi Dr , thanks so much for your informative vedio, Regarding the normal distribution of the data , what should we do if our data not normally distributed and I need to run EFA?

[@VahidAryadoust]

I suggest you can start by removing a few outliers, based on their Mahalanobis's distance. You can watch the regression and ANCOVA videos I have made, which show how to compute the Mahalanobis's distance.

[@talzabidi1569]

@@VahidAryadoust Thanks so much DR for your response , let me go to the worse case which is what about if I remove the outliers and still not normal distributed what you advise ? my second question if I need to use EFA for construct validity we should only present the output from SPSS? or we have to add some staff?

[@VahidAryadoust]

@@talzabidi1569 You might want to read more about it. To begin with, maybe look for FACTOR website, and use the FACTOR free software. It will provide you with more options than SPSS.

[@entrepreneurshipandfinance7249]

Hello, Respected Sir, EFA of 79 line items scale shows their are 15 factors but all values are loaded in first column. i dont know why.... it means there is only one dimension of scale ? but i does not seem logical. kindly guide me how to handle this.... thanks in anticipation

[@VahidAryadoust]

Check the amount of variance explained by the first factor and eigenvalues, as explained in the video. If there is only one factor, so be it.

[@nufitube3026]

HOW I CAN ANALSIS FOOD SECURITY BY SPSS?

[@danmungai555]

how do i know which variables to use in the next analysis after reduction, what has been reduced?

[@VahidAryadoust]

If the loading coefficient falls between 0.3 and 0.8, you can use them.

[@danmungai555]

@@VahidAryadoust Thank you

[@hmahadev2]

Nice video, but what after have 3 factors, what needs to be done afterwards. How do we interpret the results for each factor? Do we carry out MLR for each factor, then what is the dependent variable?

[@VahidAryadoust]

You should read the content of the items that load on each factor, and decide on a label for each factor based on the content. Technically, this will mark the end of factor analysis process.

[@hmahadev2]

@@VahidAryadoust Thank you so much

[@brantazconflix3873]

Can i collaborate EFA with Cross tabulation to extend which variable are the most appropriate among other items? Or it should be based on EFA surrogate variable?

[@VahidAryadoust]

Sorry I am not sure if I understand the question. But EFA would be sufficient to determine what variables to keep in your analysis.

[@brantazconflix3873]

@@VahidAryadoust basically, EFA collecting items into appropriate structured factor. Then, we can name the factor by its items grouped criteria. After that, can i add cross tabulation analysis for each items in a factor to make better understanding which item have more impact in a factor? For example: I've done the EFA and named the factor solution. The factor consist 3 factor and 3-5 item in each factor. My study is inpact of tourism. And my factor contain : Factor 1: economic impact (5 items) Factor 2: socio-cultural impact (4 items) Factor 3: environment impact (3 items) Can a add cross tabulation analysis which contain how many people select very agree to totally disagree to create better understanding the most direct impact felt by resident in the items for each factor??

[@VahidAryadoust]

@@brantazconflix3873 You can include additional information, if you feel it is necessary for a purpose. But that is not necessary in EFA.

[@brantazconflix3873]

@@VahidAryadoust ok. Thankyou so much 👍

[@syedabdullah5784]

Hello Sir, Could you please provide an overview of Factor anlaysis on Journal article especially empirical one's? I will share the research paper with you. Please provide an email of yours. I shall be highly obliged for your response.

[@marufahmed3806]

I used 5 point likert scale data for factor analysis.. 1= strongly disagree, 2= disagree, 0=neutral,3= agree, 4= strongly agree...is it okk???

[@VahidAryadoust]

Yes, it is.

[@elepsaelena4487]

Hello! Thank you so much for your video. It helped me a lot. I do have a question regarding my research. If for the table called Total Variance Explained appears one only component with all the numbers for Extraction Sums of Squared Loadings and for the Rotated Factor Matrix it says "Only one factor was extracted. The solution cannot be rotated", what does that mean? I tried to figure it out, but I couldn't and in 10 hours I have the exam.

[@VahidAryadoust]

It probably indicates your scale/questionnaire is unidimensional and the variance in the data cannot be decomposed. That is, there is only one factor in your data.