Statistics question

[thenewguy]

Quote from bozwood:

Actually, I am trying to do it in excel. Any ideas?

Thanks

I've done this recently... I'll see if I can find the sheet for you tonight.

- The New Guy

 

[kilbasa2]

Quote from bozwood:

Is there a way to normalize the slope of the linear regression of price data so that different units from different stocks/commodities can be compared?

Calculate the slope in absolute units then divide by the ATR (average true range) or a similar measure of volatility.

 

[horribilicus]

(Linear Regression Slope) gives you points per day.

100*((Linear Regression Slope) / (today's close)) gives you percent per day. Probably this is what you want if you are comparing two stocks.

(Linear Regression Slope) / (5 day EMA of True Range) gives you ATR's per day. Probably this is what you want if you are comparing two commodities, especially when using backadjusted continuous contracts.

 

[Bsulli]

Quote from bozwood:

Is there a way to normalize the slope of the linear regression of price data so that different units from different stocks/commodities can be compared?

Thanks in advance,
Jonathan

This is fairly close to what your looking for.
http://gummy-stuff.org/raff-regression.htm

or look around is site, tons of stuff. Gummy Rules!

http://gummy-stuff.org/

all spreadsheets can be downloaded for free.

 

[thenewguy]

http://phoenix.phys.clemson.edu/tutorials/excel/regression.html

That may help... to do it in percentage, just calculate the percentage change in the last column and use that rather than the $ change.

- The New Guy

 

[Bernoulli]

Use the log of the price series rather than the price series itself.

That way the slope of the linear regression equals (Δprice/price)/Δtime rather than Δprice/Δtime.

 

[bozwood]

Thanks everyone for the suggestions.