# What should I do to perform a confirmatory factor analysis using SPSS?

I'm trying to perform a confirmatory factor analysis using SPSS 19. I have a 240-item test, and, according to the initial model and other authors, I must obtain 24 factors. But, when I perform the factor analysis, I obtain 58 factors. What should I do to extract only 24 factors?

## Popular Answers

Rakesh Pandey· Banaras Hindu UniversityIf one does an EFA and obtains same number of factors then there is a need to assess the factor structure congruence. And if the index of factor structure congruence is high then one may conclude that the earlier reported factor structure may be generalized across different sample (population). This procedure may be termed as cross-validation of the factor structure and such procedure of factor analysis may be classified at best as inferential exploratory factor analysis and not the confirmatory factor analysis.

Confirmatory factor analysis is is a hypothesis testing approach to factor analysis where one defines the factor structure apriori and using the structural equation modeling approach how far the data fits with this predefined factor structure. If this is your objective then as suggested by others you will have to use special software like AMOS, LISREL, EQS etc.

However, if your objective is to do a cross validation of earlier reported factor structure then EFA is advised. Use any one method of EFA (preferably the one used in earlier research) and then use some tests to determine the number of factors to retain for subsequent rotation. It would be better if one adopts several criteria to decide the number of factors such as scree plot, parallel analysis, Velicer's MAP. C-Hull method etc.

If you get same number of factors as obtained in previous research even then you can not say that you have arrived at a similar factor structure unless you demonstrate that all the factors have pattern of factor loadings similar to the earlier one. This can be done by computing factor structure congruence that has been nicely described in the Book on factor analysis by Gorsuch.

Like many others, however, I also doubt that the 24 factors may not represent first order factor related to one second order latent construct. Rather, they may be related to several second order factors. And if this is the case then you will have to follow an entirely different approach to conduct the factor analysis.

Hope this helps.

Diogenes de Souza Bido· Universidade Presbiteriana MackenzieSPSS just run exploratory factor analysis (with principal components or common factor extraction).

AMOS is the package of SPSS that run confirmatory factor analysis, similar to LISREL, EQS and others.

If you do not have expertise with AMOS, LISREL…, one idea is:

- run one exploratory factor analysis for each original dimension (extracting just one factor)

- assess the convergent validity (factor loading, communality and variance extracted)

- assess the reliability (Cronbach’s alpha or composite reliability).

- you also could save the factor scores for each dimension and compute the correlation between them, to assess the discriminant validity.

Best regards,

Bido

## All Answers (96)

Paul Ingram· University of KansasAnother option, as Diogenes Bido mentioned, is that you may also just choose to do an EFA on each independent factor. This is particularly useful in structures requiring oblique rotation because it allows for the covariance between construct areas to be minimized (a good reference for this in personality instruments is Gignac, Bates, & Lang, 2007 although they use SEM instead of EFA).

Also, depending on how you determine the threshold for inclusion of a factor, you may be conceptualizing the number of extracted factors. I am not sure if you have used Monte Carlo to take into account the sample size and number of items in order to minimize noise/extra extractions (one of the risks of EFA). If not, I would highly recommend doing that and there are some fantastic programs for it that are free/easy to use.I would advise against just using the Scree plot or taking for granted what SPSS says are extracted factors above a 1.0 eigenvalue.

Best of luck.

Laura A King· University of MissouriJohn Onyango· University of Miamihttp://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=5&ved=0CFkQFjAE&url=http%3A%2F%2Fwww.ucdenver.edu%2Facademics%2Fcolleges%2Fnursing%2FDocuments%2FPDF%2FFactorAnalysisHowTo.pdf&ei=pRCOUsCEKouEkQet0IGACQ&usg=AFQjCNEXw66VNh2kfl9sgGXYj9MsK5EtVg&sig2=GZkIAFB7E60WavjobJdipw&bvm=bv.56988011,d.eW0&cad=rja

Aftab Hameed Memon· Quaid-e-Awam University of Engineering, Science and TechnologyMary I. Bresnahan· Michigan State UniversityMarion I. Van den Heuvel· Wayne State UniversityIndeed. SPSS only runs explanatory factor analysis. I would recommend using AMOS if you don't like programming yourself (syntax) as you can 'draw' the models you want to test.

Lazar Tenjovic· University of Belgrade, Belgrade, SerbiaIf you vish to do factor analysis in SPSS and limit the number of factors you should click Extraction, then tick option Number of Factors and in blank field enter 24.

You then have to calculate Tucker coefficients of congruence to find out similarity of your factors with the factors obtained by other authors.

BUT, as it has already told by many respondents to your question it is not Confirmatory factor analysis "in the proper meaning of that term". It will be in some sense "confirmatory" using of Exploratory factor analysis. If you wish to do proper CFA you should use (as it has already told by meny respondents)specialised software for CFA , like AMOS, EQS and similar packages.

Qingshan Zhao· Chinese Academy of SciencesJuan José Igartua· Universidad de SalamancaMary I. Bresnahan· Michigan State UniversityJosé Soriano Pastor· University of ValenciaYou can slect the number of factors to retain in exploratory with not select eigen=1

Steve Sizmur· Picker Institute EuropeDeletedKevin C Frissell· WestatBernhard Piskernik· University of ViennaI would like to add that you probably will have a hard time to achieve reasonable fit for such a highly multi-factored structure (24 factors) with analysis on item-level and as much as Ø 10 items/factor. Using WLMSV (I know it’s included in MPLUS) instead of the standard ML-estimation might help a little if you are using likert-style items.

Best of luck!

Muhammad-Bashir Owolalbi Yusuf· Universiti Utara Malaysia2. The number of indicator you are using are too many.

3. If what you are doing is exploratory study, I will recommend you use one of the PLS software, especially smart PLS.

4. I will recommend you see an expert close to you on the use of these statistical softwares.

5. If you are in Malaysia then I may be of help to you.

Wish you the best.

.

Mary I. Bresnahan· Michigan State UniversityGary L. Canivez· Eastern Illinois UniversityKrisdaporn Rujithamrongkul· Rangsit UniversityGarumma Tolu Feyissa· Jimma UniversityAda H Zohar· Ruppin Academic CenterRahul Pratap Singh Kaurav· Prestige Institute Of ManagementOtherwise doing it with AMOS or PLS or any other one would take a long time in drawing only.

Preety Awasthi· NITIE-National Institute of Industrial EngineeringFactor Analysis is a data dependent technique and highly related to the reliability and validity of the questionnaire. Did you check the reliability and validity of the questionnaire items you are using? Many a time the questionnaire developed in particular settings doesn't fit well in other settings. Besides this, one should always check these two parameters to ensure that the questionnaire items fulfill the requirement if in case they are adapted from the original resource.

Moreover I have a doubt whether you have taken sample from same geographical region or from a different one (many external factors related to the different geographical regions like culture, societal values etc may impact the response) . Further the responses change over a period of time even if the data is collected from the same sample (sometimes due to bias or may be due to increase in knowledge/awareness about the subject matter).

Next point is related to data itself. It seems that your data has a lot of disturbance. I suggest you to check the variability of data and variance. It may be possible that it has few outliers also.

Furthermore only reusing the questionnaire items and running factor analysis cannot be termed as confirmatory factor analysis. If you have already ran exploratory factor analysis and obtained 24 factors then fine you can move on to confirmatory factor analysis using either AMOS, LISREL, MPLUS or PLS. Before conducting the CFA please check the possibility of second or third order factors in your item set. At last if you have not conducted exploratory factor analysis yourself then first obtain a proper structure and then attempt CFA.

Best Wishes

Grace Rebekah· Christian Medical College VelloreCharles Philip Gabel· Coolum PhysiotherapyRotation with Vairmax or orthagonal

Maria Tejerina-Arreal or Tejerina-Allen· University of MurciaGrace Rebekah· Christian Medical College VelloreGideon P. De Bruin· University of JohannesburgGrace Rebekah· Christian Medical College VellorePl find the following document .

Alireza Pour-Aboughadareh· Imam Khomeini International UniversityThe pathway for this analysis:

Data reduction / factor analysis

Best regard

Saiyidi Mat Roni· Edith Cowan University2.0 54 instead of 28 factors - share with our community here your scree plot and Eigenvalue. That shall give us some clues whether that 54-factor solution is generally acceptable. Governed by a good theory, you can 'force' SPSS to extract 28 factors.

3.0 AMOS is normally bundled with SPSS. You can use this app to run CFA. Given that your preliminary extraction yields 54 instead of predicted 28 factors, my concern is your data points are widely scattered which can create identification issue (your model cannot converge) in AMOS.

Mary I. Bresnahan· Michigan State UniversityYael Fisher· Achva Academic CollegePaulo Matos Graça Ramos· Universidade Fernando Pessoa / Universidade Católica ESB / Universidade LusíadaYael Fisher· Achva Academic CollegeMary I. Bresnahan· Michigan State UniversityYuehai Xiao· New York UniversityHello researchers,

I have a very basic question :

how to run exploratory factor analysis in SPSS?

Thanks.

Mary I. Bresnahan· Michigan State UniversityYou have to enter all the items from your SPSS data file that you are factoring. There are several excellent tutorials for how to run SPSS EFA on YOUTUBE which show you step by step how to do this and then how to interpret your results.

Ada H Zohar· Ruppin Academic Centerthere is no way to run CFA within SPSS. However you can run EFA and restrict the number of factors to be extracted to 24 and see if you reach a satisfactory solution. To run CFA you can use AMOS (which belongs to the same company as SPSS but requires a separate and expensive license) or use shareware like R.

Mary I. Bresnahan· Michigan State UniversityThere is a free student download version of LISREL but it handles limited factors.

Carolyn M. Youssef-Morgan· Bellevue Universitythere is a free 14-day trial version on AMOS, fully functional.

Phillip Karl Wood· University of Missouriefa_class is available for free, if you're looking for free software that would work. Description is here: http://cran.r-project.org/web/packages/semTools/semTools.pdf

Özkan Görgülü· Ahi Evran ÜniversitesiYou can use statistica for CFA.

Diogenes de Souza Bido· Universidade Presbiteriana Mackenzielavaan is free and easy to use:

LV =~ x1 + x2 + x3

see:

ROSSEEL, Y. The lavaan tutorial. Belgium: Ghent University, Department of Data Analysis, 2015. Disponível em: <http://lavaan.ugent.be/tutorial/tutorial.pdf>;.

lavaan_site: <http://lavaan.ugent.be/resources/teaching.html>;

Nisha AroraGreat discussion !! Please also answer my queries...

@Rakesh: I couldn't completely understand "It would be better if one adopts several criteria to decide the number of factors such as scree plot,

parallel analysis, Velicer's MAP. C-Hull method etc."For factor extraction in EFA, I am aware of the following criteria: eigen vales (generally >1), TVE (>60 %), communities (>0.5), anti-image variance diagonal elements > 0.6, less than 50 % residuals to be > 0.5, apriori knowledge/preference of #factors supported by ROL (forced EFA)

Plz explain

parallel analysis, Velicer's MAP. C-Hull method.Also,someone mentioned: SPSS can be used only for EFA (NOT CFA) except for version 19 or higher. Plz share some resource on how to use SPSS for CFA. Do I need some plug-in?@Paul Ingram: which package would you recommend for factor analysis in R, I am considering FactomineR or rcmdr

But I am getting trouble interpreting the output (it's different from SPSS). Would you please share some resource for the same. Thanks

Can you help by adding an answer?