There are many manuals. However, I would advise you this one:

Fabrigar, L. R., &Wegener, D. T. (2012). Exploratory factor analysis . OUP USA.

ISBN13: 9780199734177

There are many manuals. However, I would advise you this one:

Mulaik, S. A. (2009). Foundations of Factor Analysis, Second Edition. Chapman & Hall/CRC, 2009.

ISBN: 978-1-4200-9961-4

In the following video you will see how to store the files. Remember that, at moment FACTOR allows to manage different groups of participants, and to decide which groups are included in the factor analysis at hand. This feature is a first step towards much more interesting analyses, like factor invariance analyses. We expect to be able to develop and implement such analyses in a near future.

FACTOR allows you to define (partially) specified targets that can be used to explore if a factor model is plausible one in your data. This analysis would be the unrestricted version of a confirmatory factor analysis.

FACTOR allows you to define (partially) specified targets that can be used to explore if a factor model is plausible one in your data. Please note that RETAM can help you refine your target matrix. RETAM identifies which values in the target (the ones that you defined as free values) can be set as zero values. In addition, it is advisable to use the crossvalidation analysis (also implemented in FACTOR) in order to safely interpret the refined target.

Yes and no. Yes, because FACTOR splits your data in two subsamples if your data has at least 400 observations. And no, because FACTOR does not implement random splitting algorithm. The random splitting algorithm could produce non equivalent subsamples. Instead, FACTOR implement Solomon: a method for splitting a sample into equivalent subsamples in factor analysis.

In the video tutorial, you can learn how to use it in your analyses.

Yes, all you have to do is to load the data file as a multygroup dataset. In the video tutorial, you can learn how to use it in your analyses.

Yes, it can. Here you have an example how to prepare your data in order to load it in FACTOR and to use it to compute validity studies. The example includes three groups of participants, their responses to 16 items, and the recorded value in three external variables. The labels and reliabilities of external variables are also loaded to FACTOR.

Many rules-of-thumb has been proposed based on the experience of researchers in order to decide the size of the sample. However, simulation studies suggest that the actual size of a sample to be factor analyzed depends on the characteristics of the data at hand. SENECA carries on a simulation study based on your own data, so that the size of the sample proposed takes into account the characteristics of your data.

To calculate SENECA, you must provide a small initial sample. Please note that this must be a representative sample of the population and that the larger this initial sample is, the more accurate the advice given by SENECA will be. You must decide:

- If the factor analysis to be computed is linear (i.e., Pearson correlation matrix is analyzed) or non-linear (i.e., polychoric correlation matrix is analyzed). Polychoric correlations are usually less stables, and a large sample is usually needed in order to analyze them.

- The level of precision in the analysis. It means to decide a value for the index CRMSR: the lower the value, the more accurate the simulation study will be. We advise to use at least a value between .05 and .01.

- The number of replications to be computed. We advise at least 100. The larger the number of replications, the more accurate the simulation will be.

- If the study is purely exploratory (i.e., the number of factors is totaly unknown) or it has a confirmatory aim (i.e., the number of factors is at least suspected). If a number of factors is suspected, to provide the suspected number of factors to extract adds information to the simulation study and the sample advised is usually smaller.

In this video tutorial you can see how to compute SENECA with FACTOR.

You can learn more about SENECA in the paper:

Lorenzo-Seva, U., & Ferrando, P. J. (2024). Determining Sample Size Requirements in EFA Solutions: A Simple Empirical Proposal. Multivariate Behavioral Research, 59(5), 899-912. doi:10.1080/00273171.2024.2342324.

While FACTOR was originally developed as software for exploratory factor analysis, this update takes a significant step forward by introducing methods for confirmatory factor analysis. The new method, called Mild-Restricted CFA, addresses a critical issue in traditional CFA: the distortion of factors caused by misspecified zero loadings, which lead to overestimated factor parameters and distorted structural relations. Mild-Restricted CFA mitigates these issues, allowing for the accurate assessment of CFA models without the risk of distorted parameter estimates. The ultimate goal is to establish CFA as a robust and reliable technique for factor analysis. We invite you to watch the video in the 'FAQ & Video Tutorials' section to learn how to define the factor model to be tested.

In this example we computed MR-CFA related to a factor model with perfect indicator items (i.e., each item is related to a single factor, and the secondary factor loading values are set to zero in the hypothesized factor model). Usually, the assessment of this kind of highly restricted factor model leads to a distorted estimation of the factor parameters. However, MR-CFA can deal with such a complex model, and that R-CFA fails to do so. However, in applied research there is no need to compute MR-CFA based on such restricted factor models. If previous exploratory factor analyses (computed with different samples) suggest that some secondary loading values are actually meaningful (and they have a substantive interpretation), then a more relaxed model could be proposed in which some items are not expected to be perfect indicators of a particular factor. In this sense, MR-CFA can also be used to assess factor models that are close to the Independent-Clusters-Basis (ICB) model proposed by McDonald.

A limitation that the new approach is that it is computationally intensive, and it takes more time than other approaches to be computed. We must add that this should not be a serious drawback, because computer technology advances so fast that the computing time that could be considered as excessive today, will be affordable in a short time.