The aim is to run a regression model for each cnum on observations denoted mon=13 through mon=72 using this no-intercept OLS model: crar=mkt_rf hml smb mom. The parameters of this model will be applied to the data for observation mon=73 to produce a residual for observation mon=73 and create a new variable for observation mon=73 called far that is that residual. Does this make sense?
Once that is done, we need to calculate far values for each observation from mon=73 through mon=132, in each case based upon the prior 60 observations. Then we need to calculate residuals for each cnum starting with mon=73 and going to the last observation within the cnum.