1

The textbook I'm currently reading says that the quartimax rotation in the factor analysis maximizes the rows' variance in the loading matrix.

In order to do that, it says, the quartimax rotation maximizes Q, described below. (row no i and col no j).

But about that Q, isn't it the same even if you switch the first sigma with the second one?

If it's the same, isn't the size of Q affected by both the columns and rows' variance?

I really don't understand why quartimax rotation, using Q, maximizes only the variance of rows.

I would really appreciate any help!

enter image description here

No Ru
  • 143
  • 1
    Incorrect. Not maximizing row variances but maximizes row simplicity. I.e., ideally each row (variable) having only one big loading. Power 4 is to maximize kurtosis, that is, produce fat tail of small loadings. – ttnphns Jun 18 '22 at 00:10
  • 1
    Quartimax maximizes variances of squared loadings in rows. It can be shown that this is mathematically equivalent to the simplification noted in the previous comment. – ttnphns Jun 19 '22 at 23:03
  • Thx for comment. You know where I can see that demonstration? – No Ru Jun 20 '22 at 06:41
  • 1
    In classic books on Factor analysis. Such as Harman's or Mulaik's – ttnphns Jun 20 '22 at 06:43
  • Please check this updated answer https://stats.stackexchange.com/a/185245/3277 – ttnphns Jul 12 '22 at 23:24

0 Answers0