Principal Component Analysis is a statistical technique used to reduce and aggregate variables into components for better understanding and simpler explanation.
Being a statistical technique, it can be applied once the minimum criteria and conditions for use are met.
1. Regression: PCA is applied, for example in Near-Infrared Spectroscopy (NIR) to build robust and precise prediction models for the quantification of relevant components in forage or manure. It is easier to calibrate parameters like protein, moisture or phosphat with low dimensional input data. The NIR spectra are typically high dimensional due to the large number of meassurement points in the regarding wavelenght spectrum (~800 to 2000nm, every 2nm there is reflexion meassurement -> 600 variables). In PCA: reduction to something between 3-12 PCA components ).
2. Classification/Clustering: In principle the same procedure (NIR) as with regression with the aim of recognizing clusters and classes for better explanation and better understanding. For example, the reduction to 2 or 3 dimensions, in which you can, for example, identify the effects of categorical variables (harvest year, sample type, ...).
To find the most important factors in a linear model, remove the least significant variables, and get a ranking of the importance of the factors under study. In your field, if you find a dependent variable changing with a linear relationship with respect to some independent variables, you can implement this technique.