Really depends what you are looking at, my first question is always what is the question you want to answer before you apply any statistical analysis. Also start with basic stats on the data before moving up to PCA. Look at correlations for example.

For PCA, what is the question you wish to answer, do you have multiple samples and target compounds? If the compounds are adjusted to extraction volume use that. If the data is collected in full-scan and samples have been aligned and deconvoluted properly and you have adjusted to sample volume if appropriate then use peak area. With your PCA consider if you wish to transform the data such as log transforming. Also with the PCA output, the loading plot is as important as the separation as it indicates which features drive the groupings observed.

This type of analysis does not make any sense, esp when highly variable data are suggested to be used. Statistical analysis of variables such as Rt and Area only relate to apparent concentrations taken from one method/analysis. As Gary asks, what do you hope to ask or understand? Goals are key to evaluating your question(s) and should be clear before you start. Anyone can apply statistics to a set of numbers, but that does not mean that any results obtained from doing so would have value or even be scientific. You have not provided any context about why you selected the exact data you did for analysis and why (is it all from the same LC/MS method run at at the same time or from different methods? What type of MS data are you using? Are you aware of the fact that MS detectors are not universal detectors and can easily show results that are not real, or even miss the compounds entirely? That the system is completely dependent on the knowledge and training of the operator and not the machine itself?).

Asking such a generalized question suggests that you may not have researched the technique used or the applicability of the data fields you have suggested (Rt and Area) as being useful.