Identifying the best and comprehensive tool for analyzing data and factors is an important issue in recent years. As an expert, please write your comment in this context including the name of tool and also significant reasons for your selection.
Best regards.
No single program is 'more suitable'. It depends on what you want and prefer. I use SPSS and R regularly. SPSS to perform bulk and common procedures. So data management and communicating with ACCESS or EXCEL [ODBC]. Common procedures include multilevel analyses as well. For procedure not in SPSS, I use R. For instance IRT (all types mentioned in Mair (2018) Modern Psychometrics with R), SEM, network analysis and text mining I will do in R. I use other programs as well when they have procedures and techniques not available in R. But if you want to call yourself a statistician you should be able to work with a general program (SPSS, SAS or STATA) and R.
No single program is 'more suitable'. It depends on what you want and prefer. I use SPSS and R regularly. SPSS to perform bulk and common procedures. So data management and communicating with ACCESS or EXCEL [ODBC]. Common procedures include multilevel analyses as well. For procedure not in SPSS, I use R. For instance IRT (all types mentioned in Mair (2018) Modern Psychometrics with R), SEM, network analysis and text mining I will do in R. I use other programs as well when they have procedures and techniques not available in R. But if you want to call yourself a statistician you should be able to work with a general program (SPSS, SAS or STATA) and R.
Any commercial or open access software will do, in combination if not.
Hope this helps
Matt
It depends on your research goals. However, R is free and mostly used.
Something important to consider is if you are going to work with Exploratory Factor Analysis (EFA) or Confirmatory Factor Analysis (CFA).
For the former, SPSS works pretty well, has the graphical interface (no need for using syntax, at least for basic analysis) which can be useful for a lot of people.
For CFA, AMOS is one of the easiest program to use (so far I have tried AMOS, LISREL and Mplus), you can define your model graphically, which helps in the understanding of the model for people without deep statistical knowledge. Another option for CFA can be Mplus, which is far more versatile than AMOS, but you have to work directly with the syntax. If you are going to analyze ordinal variables, do exploratory structural equation modeling (ESEM) or use robust estimators like Standardized Root Mean Square Residual (SRMR) for example, Mplus is far better than AMOS.
All these softwares are paid, something also to consider.
I don't work with R but as far as I know it should be capable of doing all these things and even more, but the learning process is harder.
As our colleagues have said, the best tool depend of which work you're going to develop. If you're just starting I would say that the SPSS/AMOS combination its a good one. But in the end, when you start understanding the concepts, the difference working with one or another software isn't that much.
Regards,
V
R software is one of the best softwares in which one can perform any kind of quantitative or qualitative analysis.This software has stats packages which can run the normal uni-variate and bi-variate tests like Chi square,T-test etc.Also on R software one can do SEM and CFA and EFA. Even qualitative analysis for which one generally uses software like NVIVO can be performed on R.It is also a free open source software.The only problem would be that R is very vast and one needs to know the right package to be used in R.
Agree with Peter Moorer on Mair's book. It covers several latent variable approaches in R (the Psychometrics in the title might put off non-educ/psych folks). I won't comment on the general value of SPSS and R; if the software does what you want and if it is what your lab uses, use it (unless you just feel like using a different package). What I will say specific to latent variable modeling is that you often want to run analyses or do something on some part of the output (the loadings, the Hessian, the correlations, etc.). This is easier with an object oriented tool, like R, than with non-object oriented tools. I am sure you can write the macros in SPSS to this, but once you start writing macros the user-friendliness that many people report for SPSS no longer exists (and at least at this level, I find R more intuitive than the macro language).
Has anyone on this thread ever used CEFA (https://www.oit.va.gov/Services/TRM/ToolPage.aspx?tid=8305). I used to use for factor analysis, and it had options not available elsewhere. It doesn't appear actively maintained though, and with different R packages doing so much probably less need for it. But shout out to Browne for this package.
Different software had different strengths and weakness.
I'm not sure I think R is necessary "the best" or that you can do everything there. R does a lot, but not everything in it is really great. For example, the mixed models package lme4 has some notable issues and shouldn't be considered as the "definitive source" in the way that SAS PROC MiXED is. Its EFA program is rather dubious but lavaan has really taken off and is under very active development.
All programs have their areas of strength and lack thereof. SPSS is good at data management and cleaning and they've implemented a number of things recently. However, last time I looked its factor analysis suite is rather out of date compared to what is considered good practice. For instance, to my knowledge it can't handle ordinal data using an ordinal estimator. Stata has quite a lot of useful things implemented, such as a pretty fully featured generalized structural equations facility that can fit many different models but you'd need to do some work to handle EFA in that framework.
Regarding CEFA, sadly Michael Browne passed away in fall of 2018. He was a true gentleman and is sorely missed by those of us who knew him. The real work he did is represented in how thoroughly he documented the EFA literature both with his articles and as implemented in CEFA, which can be used to benchmark newer programs.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software