Principal Component Analysis (PCA) | Can't get simpler!

[@aratichoudhury9030]

This is by far the best explanation of PCA I have come across. Really helpful.

[@prosmartanalytics]

Thank you! We have posted more videos on this topic. Hope you'll find them useful too.

[@ariannevanlier9960]

If you find through pca that you need 4 components instead of the 8 variables would you then suggest for further research to only ask questions regarding those 4 components? Also, what would you say as a theoretical implication contribution that you have found those 4 components if, after doing further research a regression analysis, with those 4 components and the components show no significant results?

[@prosmartanalytics]

(I)That's right, we should dig deeper into those principal components. Please check this video kzbin.info/www/bejne/eJSYg2djq7ysr6s You might want to follow PCA hands-on videos available on this channel. (II) If the objective is to predict an outcome, we can perform regression analysis using the principal components as the independent variables.

[@prosmartanalytics]

If the original data to begin with isn't appropriate to predict the outcome. PCA won't help. PCA tries to capture a significant portion of explained variance ratio but it is unsupervised so it won't now your end objective. Therefore, the quality of the information captured in the original data is definitely important.

[@ariannevanlier9960]

@@prosmartanalytics I did a regression analysis using the components, but no significant findings regarding my components. So can say that there could not be found any relation there. However, I had a really low cronbachs alpha at first therefore did pca and found these 4 underlying components. What could I know say that is my theoratical contribution regarding this pca? To clarify, the objective is not only to predict an outcome but also to make a psychometrical scale for the items that I used the PCA on.