I was walking through your correlationv1.xls spreadsheet and what came to me is how does excel's function "correl(array1,array2) handle data streams that don't match perfectly? With your example the equity hedge and high yield have matching number of values, but what if I wanted to see what the correlation is between the telecommunication & Health Care industries and I was going to take the equity prices from both industries for a 1 yr span and find the correlation. As you know each industry will have different amounts of companies, so how would excel handle it or maybe the better question is how best should we handle this scenario?
Correlation Modeling
- Thread starter idalton
- Start date
I *think* maybe you have one too many dimensions and need to clarify your X & Y down from three "dimensions" (time, company, industry). The correlation is a standardized covariance and, to my knowledge, you will need to be able to plot your dataset in 2 (X vs Y) dimensions, as the covariance is a cross product. So, every datapoint that gets included, by definition is a pair: the vectors (range sizes) must equal. If your X-axis so to speak is monthly telecom returns, then maybe (?) your Y-axis is monthly healthcare. In which case, you will summarize your companies into the monthly industry dataset (e.g., average) and in the aggregate, the disparity will disappear. My opionion is that maybe you just want to visualize the definition of the X versus the Y and figure out how your third dimension fits in.
Re: Excel, I am not sure, i didn't know it accepts mismatched arrays. My best guess is that it ignores the extras (it won't treat them like zeros, that would be wrong). I'll bet you get same answer if you delete the orphans. But back to the covariance, it can only be computed on pairs...interesting, I have challenge (defining X & Y) often when i build correlations!
David
Re: Excel, I am not sure, i didn't know it accepts mismatched arrays. My best guess is that it ignores the extras (it won't treat them like zeros, that would be wrong). I'll bet you get same answer if you delete the orphans. But back to the covariance, it can only be computed on pairs...interesting, I have challenge (defining X & Y) often when i build correlations!
David
Thank you for your response. I just wanted to take the example you gave in the spreadsheet and kind of put a different spin to it and see how it should be handled.
I want to tell you your approach to helping people grasp the difficult concepts within the FRM exam utilizing the screencasts and spreadsheets is a brilliant method.
I want to tell you your approach to helping people grasp the difficult concepts within the FRM exam utilizing the screencasts and spreadsheets is a brilliant method.
Similar threads
