statistics - guidance

[Duke Carey]

Assume you are trying to create a proxy for an indiviual stock from 30 to 50
of alternate stocks in the same industry.

With 2 years history of daily closing prices available for all the alternate
stocks, what process/analysis would you follow to identify the basket of 3-10
alternates that have most closely tracked the target stock's performance over
that time period?

Do daily closing prices add any significant accuracy to the analysis over,
say, using each Friday's closing price instead?

Any advice gladly accepted.

Duke

 

[Sean Timmons]

Well, I'd say any time you can have more data is better. But that's for you
to decide.

Wouldn't it just be easiest to have it like so...

Table A shows all stocks with daily values

Table B merely compares day by day from your stock to their stock =A2-$Z2,
=A3-$Z3, paste across and down the table.

Sum up the results to the right to get total discrepancy

Use a countif to find number of times the value is off by, say, 3.

The stock with the least number of discrepancies and the lowest total
difference would probably be your match. Make sense?

 

[Duke Carey]

Sean - thanks for the suggestion, but for diversity's sake I need to use a
portfolio approach - that is, several stocks that in the aggregate perform
most like the target stock. If you use a single stock and something causes
its value to vary widely relative to your 'benchmark,' then by definition
you've chosen a bad proxy. Using several stocks should minimize that risk
though, admittedly, something good or bad might happen to the 'benchmark'
stock too that the portfolio wouldn't mimic.