We released a change today that is especially useful when testing a single factor using the Rank Performance page. The 'Maximum Number of NAs' setting was added to the Rank Performance page and is also available in the API.
Description
NAs can cause the lower buckets to have zero stocks, resulting in non-uniform distribution of stocks in each bucket. You can achieve a uniform distribution by eliminating stocks that exceed a certain number of NAs ("Max NAs"). The setting is used like this:
On each reconstitution date:
Count how many input nodes have missing values (NA).
If the number of NAs for that stock is greater than "Max NAs", remove the stock from that reconstitution date.
Example: If "Max NAs" is 0, then a single NA in any input node removes the stock from the ranking results for that date.
Notes:
This filter is applied after ranking but before bucket assignment.
Bucket sizes will usually stay balanced across buckets unless the Min Price or Min Liquidity filters eliminate additional stocks.
This doesn't work for me. Let's say I have a ranking system with 20 nodes, some of which are composite and others of which are conditional. I want to test one particular node so I've set all the other weights to 0. But the number of NAs seems to include all nodes that are weighted 0. I have no idea how to figure out how to exclude only the stocks that are NA on the node that is weighted 100. Can this be changed so that nodes with weight = 0 aren't counted? Then I can just use 0 in the box. Right now even if I put 20 or 25 in the box all the buckets are empty.
It is counting the 0% weights as NA. I'll ask for the 0% weights to be ignored including composites and conditionals. And it currently seems to be double counting the 0 weights. So for example if you have 4 factors and 3 have 0% weight, it only returns data if max NA is set to 6 or more.
