There are 2 more confusions from the economic forecast section:
Shrinkage estimator, okay, it’s a wighted average, its good for small sample and it removes the extreme value. But after careful reading the material, I still really don’t understand how exactly this thing works.
Using ex post to forecast ex ante event: cause decrease of std and overshoot return.
- As you said it’s a simple weighted average of a parameter like “std dev” or “COV” where you basically get two different results and you weigh them e.g.
“We use the results of two different suppliers to calculate the covariance as the a weighted average of each covariance estimate.”
- In terms of statistical outputs you understimate the chances that your investiments might be a losing bet (understimate the risk) means that you have more chances that your returns will be higher than the actual ones (overestimate returns) (ref. book “Managing investment portfolios a dynamic process”).
Personally I think that he most difficult topic found among the pages of the above topics is the target cov matrix calculation (which comes from the multifactor model) where you are basically asked to calculate either :
A) the VAR of a market through its sensitivities;
B) the COV w/o correlation given with a TARGET COV MATRIX, that’s a bad beast.