Hi Mark,

The regression_length is the number of days over which the regression is computed. If you want the beta value over the last year, you should set this value to something like 252.

The returns_length is the number of days over which your returns are computed. If you want day-to-day returns, you should use a returns_length of 2. If you want rolling 5-day returns, you should use a returns_length of 5.

You should get a single beta value regardless of what you set the parameters to be.

Does this help?

Thank you!

That helps a lot

Hi Jamie,

Noob question, but why would you set the regression_length to 252 for the past year. Wouldn't it be 365?

Thanks for the help,
Thomas

Also, I'm trying to calculate the 100 day Beta like this, but it's not returning to me a floating point:

regression = RollingLinearRegressionOfReturns(  
    target=symbol('SPY'),  
    returns_length=2,  
    regression_length=100  
)

beta = regression.beta  
print(beta)

Instead it returns this in the log:

RecarrayField((RollingLinearRegressionOfReturns((Returns((USEquityPricing.close::float64,), window_length=2), Slice(Returns, column=Equity(8554 [SPY]))), window_length=100),), window_length=0)

Am I doing this correctly?

Hi Thomas,

You would set the length to 252 for the past year because there are ~252 business days in a calendar year. We do not have data outside of market hours.

For the second question, Pipeline factors aren't computed until you run your pipeline. If you haven't seen it already, I would recommend going through the Pipeline Tutorial. Essentially, you will need to add your regression factor to a pipeline, run the pipeline, and then look at the output.

Let me know if this helps.

Hey Jamie,

Thanks so much for your reply. I got it working calculating the beta using factors, but my question is, when I'm looping through stocks and trying to access the beta value, I'm not able to do something like this:

for stock in context.security_list:  
    stock.beta

How should I loop through stocks while also being able to access the respective beta values? Is this the only way to do it? I feel like it's kinda messy :/

for stock in context.security_list:  
    context.output.loc[stock].beta

Thanks for the help!

Hi everyone,

Thomas, I think the answer to your question is below. In this case, I was using 1/beta to weight the order target; however, the idea of the loop is the same:

    for security in context.security_list:  
        if data.can_trade(security):  
            order_target_percent(security,  (1/context.output.get_value(security,'beta')))  

You just have to be sure to attach the beta to the Pipeline when you build it. Does this make sense?

Does anyone know if its possible to offset this function? That is, I would like to do a rolling linear regression starting at year -2 and going to year -1. Thank you in advance for the help!

Eric