If i have let's say a sample of 30 companies, for each company i have different financial ratios and indicators for example some profitability ratios (roe, roa, eps, profit/sales...) and some financing indicators (equity/total liabilities, debt/equity...). Is it okay to use PCA with no rotation just to show two components of this data?
What's the objective of your study? What are your specific research questions? What do you want to learn about these companies, or, what do you want to be able to conclude about the relationships among the financial metrics?
If you can't articulate these, then any analysis you carry out will not be very illuminating.
Ok let me put it this way, I wan't to show which pharmaceutical company is most profitable one (which was the most profitable one in 1999 actually, since the data is from 1999) in general, so i was just wondering if doing PCA on different profitability ratios is the right choice. And I don't want an illuminating analysis, just a simple demonstration of PCA
PCA won't tell you which company is most profitable.
PCA will tell you which variables tend to "clump" together (in other words, which groups of variables or combinations of variables measure the same "concept" or "construct").
If all of the variables are related to profit in some way, then all PCA will do is confirm for you that they all "clump" together in a general construct.
If all you want to do is show which company was the most profitable, then you'll need to construct an "index" of profitability based on a combination of the financial metrics that you previously mentioned. Whichever company has the highest index is the most profitable...
Not being a financial person, I can't help much there --> This is where you'll need to do some financial research on the metrics you listed, see where they've been used before, either separately or in combination, how "profitability" is measured, how profitability is compared between two companies, etc.
If you're lucky, maybe someone has already conducted some research on profitability indices, and you can take advantage of their findings...
Google Scholar may be a good starting point, or maybe a university library....
Originally Posted by tommy
Can i also show those saved factor scores graphically somehow?