Post Your Answer

Is there a way to confirm the different measures of a single variable while doing Principal Component Analysis (PCA) on SPSS?

3 years ago in Variance By Reeta Khanna

Variables

Hi everyone, I intend to show that the items which are used for the aptitude measure are valid in the preliminary analysis of my project and that they can be scored in a single factor after the computation.

Please know that these items are taken from other studies which proved them valid already. I want to add it in my research to be double-sure as it can be proved as a learning option even when it is not the main purpose of the study.

I have been guided to do a PCA and report the percentage mentioned in the “total variance explained” calculated by adding explained % of variance listed in the first row of components.

Now, I am confused as to why only the first row’s percentage needs to be reported. Also, what is the purpose of reporting the total variance of just one component?

The written output example is given like:

By finding out the average of item score given as x (reliability: cronbach’s alphas), the Attitude Factor Score (explained percentage of the variance analysed by considered single-factor PCA) is computed.

I know this step as it has been taught; however, I want to know why in order to understand it. If anyone can help me with the explanation or share some research papers, it would be of great help.

All Answers (3 Answers In All)

By Maninder Answered 3 years ago

Hello, Well, I would say that it is not just the percentage contributed by the first component but also the cumulative percentage of all the principal components that need to be reported while conducting a principal component analysis. It helps in evaluating the approximation of the overall functions. The principal components can be 2, 3 or 4. So, after evaluating their percentage contribution, they need to be reported as well to get the overall function.

By Patric Answered 3 years ago

Hi, your question and explanation regarding your doubt is not clear enough. However, if you are trying to evaluate the variables individually to see their potential, then for the first independent variable report the combined percentage variance of that variable.

By Sumit Batra Answered 3 years ago

Hello Reeta, I believe you are trying to prove that there is a unidimensional set of variables. I hope you already understand that to have a unidimensional set of variables would mean that the variance percentage as reported by using the first extracted component is substantial when computed with a good item component or factor for all existing items and that the subsequent factors which are extracted are minor as compared to the first defined factor. I know you have been advised to choose the PCA approach but I would recommend you to go with Factor analysis with the approach of maximum likelihood extraction or the principal axis. Also, the reason for looking at the variance computed by the first component is to prove whether it is substantial or not. To have a better understanding of the measure structure, once the declared one-factor model is ready, you can run a confirmatory factor analysis test to verify the sample. I hope this provided you with the information you wanted. If I come across any relevant study or paper, I will share it. Best regards, Sumit Batra.

Replied 3 years ago

By Reeta Khanna

Hello Sumit, thank you so much for briefing the possible ways and explaining the reason. I will look more into these approaches to have a better knowledge about them. Thank you once again.

View Related Questions