Different aggregation method within the same level

[FNRobinne]

COINr allows for different aggregation methods between levels, but it would be nice to be able to specify different methods within the same level: I might want a dimension aggregated using a arithmetic mean and the three others aggregated with a geometric mean.

The only idea I have to do that right now is to create pseudo levels:

• pseudo level 1 assigned with arithmetic mean, with parent level 3
• pseudo level 2 assigned with geometric mean, with parent level 3
• Level 3 assigned to the dimension level

Make sense? Still have to test it, but I figured you might have a different way to do it. Thanks.

[bluefoxr]

Hi, at the moment COINr doesn't support this. I can leave it on the list of possible features to add but at the moment I don't have sufficient time to implement it.

I'm not sure if your solution would work. COINr was written expecting levels to be consecutive, whereas there you have level 1 with parent at level 3. To be honest I don't know what the implications of doing that are, whether COINr would accept it or not, and whether there would be any strange consequences if you do that. So if you go that way, just check carefully what happens.

[FNRobinne]

Thank Will. Yes, I think it would be a great addition to COINr's features. I'll give it more thoughts; I started to implement the solution above, but so far returning errors, because it creates domino errors (mismatches) in the iMeta file. Will keep trying and report.

Because I only have one dimension (out of four) that needs an arithmetic mean, I can compute it separately and add it to the coin as the sole indicator of its dimension (in other words, the indicator is already an aggregate). I can run it on the original dataframe still using COINr with the selected columns and then pass this dataframe into the coin building function.