pspp-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Factor Analysis Promax


From: Andreas Hammer
Subject: Re: Factor Analysis Promax
Date: Thu, 18 Apr 2024 18:48:19 +0200
User-agent: Mozilla Thunderbird

Dear Mr. Walter.

Thank you very much.

Kind regards

Andreas Hammer

Am 15.04.2024 um 19:46 schrieb Dr. Oliver Walter:

Dear Mr Hammer,

I tested the results of factor analysis with Promax rotation in PSPP. For a sample data set of 10 items and 100 persons the results agreed, only the order of the factors were different. Hence, first it looked like if the results were different. After I put the factors in the same order, the results were the same. One factor was also inverted, hence, the loadings and correlations had to be multiplied with -1. But this is not a problem.

Kind regards,

Dr. Oliver Walter

Am 15.04.2024 um 14:09 schrieb Hammer:

Hello Mr. Walter,


thank you for your exlanation. Until yet I've never used Promax. Probably I was confused, because Promax is not part of the GUI and i was not sure, if the procedure weorks well.

Kind regards

Andras Hammer

Am 14.04.2024 um 08:05 schrieb Dr. Oliver Walter:

Hello, Mr Hammer,

Promax is an olique rotation method, i.e. the factors can be correlated after rotation. Then loadings of items on factors and correlations of items with factors are different from each other. Hence, there are two matrices of coefficients: the structure matrix, i.e. the matrix of correlations of items with factors, and the pattern matrix, i.e. the matrix of loadings of items on factors. With uncorrelated factors both matrices are the same and only one of them has to be printed. 

The sum of squared loadings is not meaningful with correlated factors. With uncorrelated factors the sum of squared loadings of a factor indicates the proportion of variance explained by a factor, but not with correlated factors.

Kind regards,

Dr. Oliver Walter

Am 13.04.2024 um 22:20 schrieb Andreas Hammer:

Hello,

The manual specifies Promax as a rotation method for the factor analysis. I cannot call up this method via the menu. If I enter it via the syntax, a factor analysis can be run - but in my dataset:

1) the sum of squared loadings is not calculated, 

2) instead of a rotated matrix I get a "pattern matrix (Mustermatrix)". Maybe it's the same?

With Varimax there is no problem.

What have I overlooked?

Thank you.


Kind regards

Andreas Hammer

                      Systeminformation
╭────────────────────────┬──────────────────────────────────╮
│Version                 │GNU pspp 2.0.0-g4c33fd            │
│Hostsystem              │x86_64-w64-mingw32                │
│Build System            │x86_64-pc-linux-gnu               │
│Lokale Sprachverzeichnis│C:\Program Files\PSPP/share/locale│
│Compilerversion         │10-win32 20210110                 │

╰────────────────────────┴──────────────────────────────────╯

I use win 11.

---------------------------------

Syntax like this:

FACTOR
    /VARIABLES= WW02_29 WW02_02 WW02_03 WW02_30 [and more items, totaly 64]
    /CRITERIA = MINEIGEN (1) ITERATE (25)
    /EXTRACTION =PAF
    /METHOD = CORRELATION
    /PLOT = EIGEN
    /PRINT = INITIAL EXTRACTION ROTATION SIG AIC KMO
    /CRITERIA = ITERATE (25)
    /ROTATION = VARIMAX.


-

reply via email to

[Prev in Thread] Current Thread [Next in Thread]