[R] Estimators for non-normal data

[Martin Maechler]

>>>>> Nika Sušac <nika.susac at gmail.com>
>>>>>     on Sat, 23 Jul 2016 19:39:48 +0200 writes:

    > Hi!  I have non-normal data (items are continuous on a
    > 9-point scale, but not normally distributed) and I want to
    > conduct cfa. Which of the estimators available in lavaan
    > do you recommend me to use?  Thanks in advance!

I think you want *robust* statistical methods then.

Robust factor analysis (if 'cfa' means something like that) is
somewhat prominent topic.
Simple approaches will already be available by using
MASS::cov.rob() for a robust covariance matrix which you then
can pass to other methods.

For more (and more modern) methods and approaches,

- for R packages, I'd recommend you consult the CRAN task view
  about robust statistical methods,
  	https://cran.r-project.org/web/views/Robust.html
  notably the section on 'Multivariate Analysis'

- for more specific help and expertise, I strongly recommend
  the dedicated  R-SIG-robust mailing list (instead of R-help),
   --> https://stat.ethz.ch/mailman/listinfo/r-sig-robust

Best regards,
Martin Maechler, ETH Zurich

    > 	[[alternative HTML version deleted]]