Code comments

[jrising]

openjournals/joss-reviews#7538

Comments on code in R:

• gaia/R/crop_calendars.R

  Line 26 in f4a0e78

  d$plant <- ifelse(d$Plant.start < d$Plant.end, ((d$Plant.start + d$Plant.end) / 2 / 30), ((d$Plant.start + d$Plant.end + 365) / 2 / 30))

  : Day-of-year divided by 30 isn't a great approximation for month, unless you're working with 360-day calendar data.

• gaia/R/crop_calendars.R

  Line 47 in f4a0e78

  # Deal with exceptions (countries by states or regions)

  : This would be less error-prone if encoded in a data file.

• gaia/R/data_aggregation.R

  Line 29 in f4a0e78

  input_file = "crop_calendar.csv"

  : If you intend for users to modify this, provide them a way to use their own filename.

• gaia/R/data_aggregation.R

  Line 51 in f4a0e78

  pattern = utils::glob2rx("*precip*rfc*"),

  : Why is only the RFC weighted data used?

• gaia/R/weighted_climate.R

  Line 98 in f4a0e78

  dplyr::left_join(mirca_harvest_area, by = c("lon", "lat")) %>%

  : This assumes a 1-to-1 mapping between region and grid cell. At the country level, this is usually fine, but would not work at higher resolutions.

• gaia/R/yield_impacts_functions.R

  Line 439 in f4a0e78

  reg$se_robust <- sandwich::vcovHC(reg, type = "HC", weights = iso)

  : The SE should also be clustered, since there will be correlation between error terms in the each country.