Fill NAs with zero in abcd

[jdhoffa]

This PR adjusts the expected behaviour of target_market_share in how it handles input abcd data when abcd$production contains NAs. Previously it would filter these rows and remove from analysis. Now, it replaces them with 0.

@jacobvjk I have a follow-up question, which is, do we want target_sda to exhibit the same behaviour? I imagine probably we do.

and @cjyetman I wanted your input here, do you think this is a reasonable thing to do? (see linked issue for more reference). I know there are, in general, fears around replacing NAs with 0s willy nilly, but I think in this particular case, we do actually expect that PAMS data with NA production values DO mean 0 production... Especially if the asset has not "turned on" yet.

Closes #423

[jdhoffa]

Note: It seems #455 has introduced a note in R CMD check... it is not complaining that r2dii.match is in "Depends" but not imported or called in the source code. We may want to revisit that...

Since this is only a NOTE, it won't appear in our automated CI/CD, but it's something worth looking into in any case.

Handling this in #461

[cjyetman]

No strong opinion here because @jdhoffa and @jacobvjk know better what needs to be done here than I do, but general thoughts...

• replacing NA with 0 can, under some circumstances, significantly change what the data means
• replacing NA with 0 at the beginning of a time series that has >0 values after it is probably mostly safe, particularly in this context
• replacing NA with 0 at the end of a time series after >0 values probably has a higher chance of being wrong

[jdhoffa]

Thanks CJ, that was the context I was hoping for! :-)

In that case, I think the gist of this PR is ok, though perhaps the implementation should be a bit more precise... @jacobvjk what do you think

[jacobvjk]

I still feel like it is reasonable safe to assume that an NA is an asset that is not actually operating at that point in time. In that case a replacement with zero is fine. I remember being told in the data herald that when no updated information on production sites are available, PAMS assumes that the level of production/capacity from the previous year is still valid.

[jacobvjk]

@jacobvjk I have a follow-up question, which is, do we want target_sda to exhibit the same behaviour? I imagine probably we do.

I am not so sure about this one. imagine you have a steel plant that is not operational in a given year (has either 0 or NA production). I would expect that steel plant to have an NA value for emission intensity. clearly the technology employed would not be a zero carbon technology even though nothing is produced. so I think you can't replace the NA with a 0 here. If you did, the exposure to this plant would be considered zero emissions and you could reduce your portfolio's emission intensity by having high financial exposure to an inactive plant.

Am I understanding your point right? @jdhoffa

[jdhoffa]

Makes sense! And agreed.

Then I think this PR is good to go

[jacobvjk]

lgtm