Principal Component Analysis

Leonardo_C

Hello,

I have just installed JASP 0.16 and I ans still learning to use it. I like because it is easy to use and have some complex statistical tests that I can understand, but i am not able to calculate myself.

I am, however, a bit confused about the tool for Principal Component Analysis for two reasons.

The first is that I was able to get only Components loadings and eigenvalues, but I have not found a way for getting object scores (i.e. the projections of observations along the principal axes) and residuals.

Is this tool totally "variable oriented" or have I missed something?

The other obscure point is that when I toggle between decomposition based on covariance matrix and correlation matrix I obseved no changes in the output (loadings and eigenvalues). This looks odd to me, because working on correlation matrix should be equivalent to use standardized (unit variance) input variables, giving them equal a priori importance.

May someone help me?

Thanks in advance.

Leonardo_C

eduard

Hi Leonardo,

I moved this discussion to JASP. There you will get an answer quicker.

Eduard

julius

Hi Leonardo,

I will move your comments to our GitHub issues page. What you mention is 1) a feature request, since you are correct, we do not offer what you seek, and 2) a bug report, because this seems to be a bug :)

julius

Here are the issues. It would help out a lot if you could take a look and let me know if anything is missing.

https://github.com/jasp-stats/jasp-issues/issues/1666

https://github.com/jasp-stats/jasp-issues/issues/1667

Leonardo_C

Dear Julius,

you reported the issues correctly at GitHub. Thanks.

I just have a doubt. In my version of JASP there is no the option "Add PCA scores to data", what version of JASP do you use? Mine is JASP 0.16.1.0.

Leonardo_C

julius

Hi Leonardo, sorry for the delay. Seems I do not get notifications from this forum in my email. The option only appears in some nightly builds and is available only in debug mode. However, the option does not have any effect so far, but I have let some people know about the problem and with any luck they might get to it for one of the next releases. Sorry I cannot give you better news :)

Leonardo_C

Thanks for explanation Julius, but I does not understand what "nightly builds" exactly means.

Leonardo_C

julius

Nightly builds are a sort of beta-version downloads that might be unstable but mirror the most recent updates to jasp.

EJ

Available here: https://static.jasp-stats.org/Nightlies/

Ghislain

Bonjour, 

Je suis utilisateur de JASP, la dernière version disponible sur internet. Pouvez-vous me dire comment générer les variables ou facteurs latents ou composites, lors d'une EFA ou ACP ? C'est à dire les faire apparaitre dans le tableau des données. Pour ensuite les utiliser comme des variables pour des régressions ou corrélations. 

Merci pour vos réponses.

GF

Ghislain

Hello, 

I am a user of JASP, the latest version available on the internet. Can you tell me how to generate the latent or composite variables or factors, during a CTA or PCA? That is, to make them appear in the data table. Then use them as variables for regressions or correlations. 

Thank you for your answers.

GF

julius

Hi Ghislain,

thank you for your question. If I understand correctly, this touches on the same subject Leonardo raised, which is to add the component scores or factor scores to the data. Unfortunately, this is not possible so far, but it is on the radar of the JASP team for future implementation. In order to use factors in regression with other factors you could also try out the SEM module, of course.

Regards,

Julius

Ghislain

Julius,

Un grand merci pour votre réponse, je vais me former à la procédure SEM et voir si cela me permet de répondre à mes attentes.

Cordialement

Ghislain

Julius,

Thank you very much for your answer, I will train myself in the SEM procedure and see if it meets my expectations.

Sincerely