Trigonelline and caffeine are found in small quantity whereas protein and fat are found in higher quantities in coffee beans. These variables are measured together from the same samples but found in different quantities in the coffee beans. Therefore should I use a correlation matrix?
Check https://stats.stackexchange.com/questions/53/pca-on-correlation-or-covariance
Hi Getachew Weldemichael ,
The general considerations of the above link are valid and finally you have to think about the purpose of your experiment.
In the end, you can choose the "best" PCA method or even other multivariate methods.
Good luck
Dear all, thanks for your response but I have one question for Dr. Roberto Molteni. The purpose of the study is to identify the variables which accounted for the lion's share of the observed variability, So what should I use?
Hi Getachew Weldemichael ,
If the dataset is the proximate analysis of many samples, usually and simplifying there are macro and micro components (variables).
The quantity of the component affects the assessment of variability; therefore, it might be useful to test the difference in results using two approaches: PCA with the correlation matrix and PCA with the covariance matrix.
Another approach could be to find similar samples in your multivariate space, e.g. with a cluster analysis, precisely to have more homogeneous groups of samples and more easily control the impact of the different variables on variability.
Otherwise, you have other approaches, e.g. Variance components, which is in its simple case https://www.itl.nist.gov/div898/handbook/prc/section4/prc44.htm
By the way, the topic might be of interest to other researchers as well, and you could ask another question a little more detailed for contributions from other researchers.
What many people seem to be asking is what is your research question? David Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Analysis
- Data Analysis
- Dataset
- Big Data