add new summarize_weighted_buildout function #141
Conversation
|
Surprisingly, all calls to join seem to use #> Joining, by = c("sector", "technology")-- library(r2dii.data)
library(r2dii.match)
#> Warning: package 'r2dii.match' was built under R version 4.0.2
devtools::load_all()
#> Loading r2dii.analysis
master <- loanbook_demo %>%
match_name(ald_demo) %>%
prioritize() %>%
join_ald_scenario(ald_demo,
scenario_demo_2020,
region_isos = region_isos_demo
)
summarize_weighted_buildout(master)
#> Joining, by = c("sector", "technology")
#> # A tibble: 138 x 4
#> sector technology year weighted_buildout
#> <chr> <chr> <int> <dbl>
#> 1 automotive electric 2020 0
#> 2 automotive electric 2021 0.000000765
#> 3 automotive electric 2022 0.00000153
#> 4 automotive electric 2023 0.00000230
#> 5 automotive electric 2024 0.00000306
#> 6 automotive electric 2025 0.00000383
#> 7 automotive electric 2026 0.00000459
#> 8 automotive hybrid 2020 0.0000528
#> 9 automotive hybrid 2021 0.0000501
#> 10 automotive hybrid 2022 0.0000474
#> # … with 128 more rowsCreated on 2020-07-23 by the reprex package (v0.3.0) -- Found the issue here. A call to |
The previous approach is too recent in rlang, and would break unless the user has a recent version of rlang.
|
@maurolepore yes it's safe to update. Im truly not sure what is causing them to mess up. |
Conflicts: tests/testthat/test-summarize_weighted_production.R
| test_that("is sensitive to `use_credit_limit`", { | ||
| data <- fake_master( | ||
| name_ald = rep(c("company a", "company b"), 2), | ||
| year = c(2020, 2020, 2021, 2021), | ||
| id_loan = c("i1", "i2", "i1", "i2"), | ||
| loan_size_credit_limit = c(20, 30, 20, 30), | ||
| production = c(10, 10, 20, 40), | ||
| ) | ||
| out2 <- data %>% | ||
| summarize_weighted_percent_change(name_ald, use_credit_limit = TRUE) | ||
| out3 <- data %>% | ||
| summarize_weighted_percent_change(name_ald, use_credit_limit = FALSE) | ||
|
|
||
| expect_false(identical(out2, out3)) | ||
| }) |
There was a problem hiding this comment.
This comes from expectations that I found outside a call to test_that(). I change the code to reflect how I would test that summarize_weighted_percent_change() is sensitive to use_credit_limit.
| # TODO: Come back to this after the methodology article is written | ||
| # test_that("with known input outputs as expected", { | ||
| # # styler: off | ||
| data <- fake_master( | ||
| name_ald = rep(c("company a", "company b"),2), | ||
| year = c(2020, 2020, 2021, 2021), | ||
| id_loan = c("i1", "i2", "i1", "i2"), | ||
| production = c( 10, 10, 20, 40), | ||
| ) | ||
| out1 <- summarize_weighted_percent_change(data, name_ald) | ||
| out1$weighted_percent_change | ||
| expect_equal(out1$weighted_percent_change, c(0,0,50,150)) | ||
|
|
||
| # Is sensitive to `use_credit_limit` | ||
| # Reversing loan_size and production outputs reverse result | ||
| data2 <- fake_master( | ||
| name_ald = rep(c("company a", "company b"),2), | ||
| year = c(2020, 2020, 2021, 2021), | ||
| id_loan = c("i1", "i2", "i1", "i2"), | ||
| loan_size_credit_limit = c( 20, 30, 20, 30), | ||
| production = c( 10, 10, 20, 40), | ||
| ) | ||
| out2 <- summarize_weighted_percent_change(data2, name_ald, use_credit_limit = TRUE) | ||
| expect_equal(out2$weighted_percent_change, c(0,0,40,180)) | ||
| # # styler: on | ||
| # }) | ||
|
|
There was a problem hiding this comment.
@jdhoffa, I'm not sure about removing this. It wasn't inside a test, nor it seems that the code and comments correspond to each other. Can you check this is really to throw away?
There was a problem hiding this comment.
ah yes, i didn't mean to remove the test entirely! i meant only to remove the TODO comment haha.
I put the test back in
| @@ -0,0 +1,61 @@ | |||
| --- | |||
| title: "Indicator Choices: Production vs. Percent Change in Production" | |||
There was a problem hiding this comment.
Is "production" enough or do we need to qualify "production" with some word to more precisely distinguish it from "percent change in production"? Without a qualifier, percent in production seems like a subgroup of production when it is instead an alternative.
Three qualifiers that come to my mind are weighted, absolute, and raw.
There was a problem hiding this comment.
I elected to add "Weighted" in every case
| We aggregate the percent-change in production for each company to the portfolio-level, by using the same loan-weighted average as above. In particular, for each loan $l_j$ to company $j$, we have: | ||
| $$ \overline{\chi_i} = \sum_j \left[ \chi_{i,j} * \dfrac{l_j}{\sum_j l_j} \right]$$ | ||
|
|
||
| It should be noted that the percent change, $\chi$, is undefined for $0$ initial production. Intuitively, this makes sense, since you would require an "infinite percent" build-out to grow to anything from 0. For this reason, any company having $0$ initial production is filtered out prior to calculating the percent change indicator. |
There was a problem hiding this comment.
Where it says "$0$ initial production" to replace
Maybe you mean "When initial production (
|
|
||
| ```{r} | ||
| # using the master dataset defined in the previous chunk: | ||
| summarize_weighted_percent_change(master) |
There was a problem hiding this comment.
I removed the calls to head() because they seem necessary for tibbles. Is there a reason you prefer head()?
There was a problem hiding this comment.
Hmmm I thought they were formatting strangely without head(), but it looks fine to me now.
There was a problem hiding this comment.
@jdhoffa, I suggested a few changes and made a few comments. If you feel it's ready, then let's merge. If not, I'm happy to review again after you incorporate my review.
|
I will merge once the checks pass! Thanks @maurolepore |
This is a WIP, and requires #136 to be merged before checks will pass. I open here as a draft just to track my work.
Remains to be done:
summarize_weighted_buildoutis very similar tosummarize_weighted_productionand I imagine there must be a more DRY way to accomplish thisRelates to #132