Add weighted_average_lat

[Sbozzolo]

And add option to add weights to average_lat,

I didn't make it default to maintain backwards compatibility, but we can make it default in the future.

Closes #38

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.62%. Comparing base (e1d62f2) to head (ce1c74a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #39      +/-   ##
==========================================
+ Coverage   94.39%   94.62%   +0.22%     
==========================================
  Files           7        7              
  Lines         357      372      +15     
==========================================
+ Hits          337      352      +15     
  Misses         20       20              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[szy21]

Thanks! Just a few comments.

[szy21]

Suggested change

	maximum(lats) >= 0.5π ||
	@warn "Detected latitudes are small. If units are radians, results will be wrong"
	maximum(abs(lats)) >= 0.5π ||
	@warn "Detected latitudes are small. If units are radians, results will be wrong"

(it is common to take e.g. southern hemisphere mean, and this way the warning wouldn't show up for that case)

[Sbozzolo]

Thank you!

[szy21]

If _reduce_over takes the average, the weights here should be divided by the average weights?

[Sbozzolo]

Right, I need to normalize the weights!

[szy21]

If you do this the average of var should be sum(var.data) instead of mean(var.data), because the sum of the weights is 1 now? Alternatively we can do weights ./= mean(weights_1d) I think.

[Sbozzolo]

Yes, thank you! I changed to mean

[szy21]

@cmschmitt519 You can use this once it's merged.

[Sbozzolo]

@szy21 can you have another look?

[szy21]

Thanks!

[Sbozzolo]

@cmschmitt519 If you need this, it is available in the most recent version (0.5.6)