I have data on a $M$ systems (say different material alloys). Each system (material) has $N$ variables (properties). I would like to correlate one variable(say material strength) of the system with a small sub-set of the variables $N_S$ (such as composition, crystal-structure etc.).

I'm not sure which analysis is more appropriate for this scenario. Do I need to use

  • factor analysis -- which I believe treats all variables as dependent variables or
  • multiple regression -- which treats one variable as dependent and others as independent variable.

Just to clarify, the variable values are not continuous i.e., for given value/range of variables, a system/systems may or may not exist.

  • $\begingroup$ why not just do a correlation matrix? $\endgroup$ – mandata Aug 13 '15 at 16:08
  • $\begingroup$ I have no background in statistics. I read a little bit online and I just posted a question based on what I learnt $\endgroup$ – WanderingMind Aug 13 '15 at 16:18
  • $\begingroup$ In that case, start with a correlation matrix, and do a graphic, too, not just numbers. It is available in most stat packages. That is good, standard exploratory practice. $\endgroup$ – mandata Aug 13 '15 at 16:20

When starting out with a data analysis, it is good practice to "look" at the data, which means poring over graphs of distributions, descriptive statistics (frequencies, proportions, means, ranges, etc), and try to get a feel for the data, without doing analysis that answers specific questions. In this case I would recommend building correlation matrices, both graphically and numerically. This will give you a sense of how the variables are related to each other, and will suggest what method(s) might be best to proceed.

| cite | improve this answer | |
  • $\begingroup$ Thank you. Correlation matrix definitely looks like a good place to start $\endgroup$ – WanderingMind Aug 20 '15 at 12:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.