Factors return a single scaler value per output. There's no way to get a time series or dataframe from a factor.
One could simply use the 'data.history method (https://www.quantopian.com/help#ide-history) to get a dataframe of prices for each security. Use the pipeline to fetch all the securities in Q500US then pass that list to the 'data.history' method. Finally, do whatever calculations one wishes with those prices. However, that approach would be slower than doing the same calculations inside the pipeline. For arbitrarily complex calculations consider writing a custom factor.
Here is an example of how one can pass log returns for the universe of securities, along with the log returns for a single security, to the 'compute' function of a custom factor.
class Mean_Diff_Log_Return_From_SPY(CustomFactor):
'''
Returns the mean of the difference of 1 day log returns from SPY.
window_length is the number of days to average. Default is a single day but can can be over-ridden.
'''
returns = Returns(window_length=2).log1p()
returns_spy = returns[symbols('SPY')]
inputs = [returns, returns_spy]
# Set window_length to number of days returns to average
window_length = 1
def compute(self, today, assets, out, returns, returns_spy):
# This is just a sample of the kind of logic one could do inside of the compute method
out[:] = (returns - returns_spy).mean(axis=0)
Attached is a notebook with this same example in action. This may be the approach to use to check for co-integration with SPY?
