This commit is contained in:
Pablo Antonio Lillo Cea 2026-05-08 10:12:33 +02:00
parent 1631adfb02
commit eb093605b3
2 changed files with 79 additions and 64 deletions

View file

@ -42,7 +42,7 @@ library(FactoMineR)
# 00-Load ----------------------------------------------------------------------
m_sample <- read_rds("data/processed/m_sample.rds")
all_vars <- setdiff(names(m_sample), c("code", "municipality"))
all_vars <- setdiff(names(m_sample), c("code", "municipality"))
# 01-Exclude genuinely uninformative variables ---------------------------------
# Removed from every part of the analysis: redundant totals, geographic areas,
@ -57,7 +57,9 @@ truly_exclude <- c(
"land_area_ha",
all_vars[str_detect(all_vars, "^type_of_land_")],
all_vars[str_detect(all_vars, "^use_of_land_")],
"forest", "open_land", "total_green_space",
"forest",
"open_land",
"total_green_space",
all_vars[str_detect(all_vars, "^inland_water")],
all_vars[str_detect(all_vars, "^seawater")],
all_vars[str_detect(all_vars, "^the_four_large")],
@ -90,23 +92,34 @@ col_sup_politics <- analysis_vars[
# rather than resident persons. Projecting them as supplementary shows how
# they relate to the population-composition space without distorting it.
col_sup_infra <- analysis_vars[
str_detect(analysis_vars,
"^number_of_(rented|tenant_owned|owner_occupied)_dwellings") |
str_detect(
analysis_vars,
"^number_of_(rented|tenant_owned|owner_occupied)_dwellings"
) |
str_detect(analysis_vars, "^number_of_registered_passenger_cars_") |
str_detect(analysis_vars, "^workplaces_") |
str_detect(analysis_vars, "^agricultural_enterprises_") |
str_detect(analysis_vars, "^livestock_") |
analysis_vars %in% c(
"sex_men", "sex_women",
"employment_by_gender_men", "employment_by_gender_women",
"number_of_inmigrations", "number_of_outmigrations",
"births", "deaths", "marriages", "divorces",
"buildings", "buildings_for_seasonal_use",
"concentrations_of_holiday_homes", "holiday_home_areas",
"social_assistance_number_of_receiver_households",
"urban_residences_proximity_to_public_green_areas_500_meters_or_less",
"number_of_localities"
)
analysis_vars %in%
c(
"sex_men",
"sex_women",
"employment_by_gender_men",
"employment_by_gender_women",
"number_of_inmigrations",
"number_of_outmigrations",
"births",
"deaths",
"marriages",
"divorces",
"buildings",
"buildings_for_seasonal_use",
"concentrations_of_holiday_homes",
"holiday_home_areas",
"social_assistance_number_of_receiver_households",
"urban_residences_proximity_to_public_green_areas_500_meters_or_less",
"number_of_localities"
)
]
col_sup_vars <- c(col_sup_edu, col_sup_politics, col_sup_infra)
@ -119,24 +132,16 @@ col_sup_vars <- c(col_sup_edu, col_sup_politics, col_sup_infra)
# compositional profiles.
active_vars <- analysis_vars[
(str_detect(analysis_vars, "^age_") |
str_detect(analysis_vars, "^education_level_of_swedish_men_") |
str_detect(analysis_vars, "^education_level_of_swedish_women_") |
str_detect(analysis_vars, "^employment_by_activity_sectors_") |
str_detect(analysis_vars, "^birth_country_")) &
str_detect(analysis_vars, "^education_level_of_swedish_men_") |
str_detect(analysis_vars, "^education_level_of_swedish_women_") |
str_detect(analysis_vars, "^employment_by_activity_sectors_") |
str_detect(analysis_vars, "^birth_country_")) &
!analysis_vars %in% col_sup_vars
]
# Everything else → post-hoc correlations with CA dimensions
outside_ca <- setdiff(analysis_vars, c(active_vars, col_sup_vars))
cat(
"Active (person-count population composition): ", length(active_vars), "\n",
"col.sup educational provision: ", length(col_sup_edu), "\n",
"col.sup political vote counts: ", length(col_sup_politics), "\n",
"col.sup infrastructure / event counts: ", length(col_sup_infra), "\n",
"Outside CA (rates / continuous / other): ", length(outside_ca), "\n"
)
# 03-Build CA matrix -----------------------------------------------------------
X <- m_sample |>
select(all_of(c(active_vars, col_sup_vars))) |>
@ -154,9 +159,6 @@ idx_sup <- seq(length(active_vars) + 1L, ncol(X))
# 04-Run CA --------------------------------------------------------------------
ca <- CA(X, ncp = 10, col.sup = idx_sup, graph = FALSE)
cat("\nEigenvalues (first 10 dimensions):\n")
print(round(ca$eig[1:10, ], 3))
contribs <- ca$col$contrib |>
as.data.frame() |>
rownames_to_column("variable")
@ -177,7 +179,11 @@ outside_data <- m_sample |>
replace_na(x, if (is.finite(m)) m else 0)
}))
posthoc_cor <- cor(ca_row_coords, outside_data, use = "pairwise.complete.obs") |>
posthoc_cor <- cor(
ca_row_coords,
outside_data,
use = "pairwise.complete.obs"
) |>
as.data.frame() |>
rownames_to_column("dimension")
@ -185,11 +191,11 @@ posthoc_cor <- cor(ca_row_coords, outside_data, use = "pairwise.complete.obs") |
write_rds(ca, "data/processed/ca_exploratory.rds")
write_rds(
list(
active = active_vars,
edu = col_sup_edu,
active = active_vars,
edu = col_sup_edu,
politics = col_sup_politics,
infra = col_sup_infra,
outside = outside_ca
infra = col_sup_infra,
outside = outside_ca
),
"data/processed/ca_var_groups.rds"
)

View file

@ -26,13 +26,13 @@ library(jsonlite)
SCB_URL <- "https://api.scb.se/OV0104/v1/doris/sv/ssd/START/UF/UF0506/UF0506B/Utbildning"
YEARS <- c("2000", "2005", "2010", "2015", "2022")
YEARS <- c("2000", "2005", "2010", "2015", "2022")
LEVELS <- as.list(as.character(1:7))
AGES <- as.list(as.character(25:64))
AGES <- as.list(as.character(25:64))
GENDERS <- list("1", "2")
# Fetch municipality codes from table metadata
meta <- fromJSON(content(GET(SCB_URL), "text", encoding = "UTF-8"))
meta <- fromJSON(content(GET(SCB_URL), "text", encoding = "UTF-8"))
munis <- Filter(\(x) nchar(x) == 4, meta$variables$values[[1]])
cat("Municipalities to fetch:", length(munis), "\n")
@ -40,23 +40,27 @@ cat("Municipalities to fetch:", length(munis), "\n")
fetch_batch <- function(muni_batch) {
query <- list(
query = list(
list(code = "Region",
selection = list(filter = "item", values = as.list(muni_batch))),
list(code = "Alder",
selection = list(filter = "item", values = AGES)),
list(code = "UtbildningsNiva",
selection = list(filter = "item", values = LEVELS)),
list(code = "Kon",
selection = list(filter = "item", values = GENDERS)),
list(code = "Tid",
selection = list(filter = "item", values = as.list(YEARS)))
list(
code = "Region",
selection = list(filter = "item", values = as.list(muni_batch))
),
list(code = "Alder", selection = list(filter = "item", values = AGES)),
list(
code = "UtbildningsNiva",
selection = list(filter = "item", values = LEVELS)
),
list(code = "Kon", selection = list(filter = "item", values = GENDERS)),
list(
code = "Tid",
selection = list(filter = "item", values = as.list(YEARS))
)
),
response = list(format = "json")
)
resp <- POST(
SCB_URL,
body = toJSON(query, auto_unbox = TRUE),
encode = "raw",
body = toJSON(query, auto_unbox = TRUE),
encode = "raw",
content_type("application/json"),
timeout(60)
)
@ -67,21 +71,30 @@ fetch_batch <- function(muni_batch) {
# key is a list of character vectors: [region, age, edu_level, gender, year]
keys <- do.call(rbind, d$key)
tibble(
code = keys[, 1],
age = as.integer(keys[, 2]),
code = keys[, 1],
age = as.integer(keys[, 2]),
edu_level = as.integer(keys[, 3]),
gender = as.integer(keys[, 4]),
year = as.integer(keys[, 5]),
n = as.integer(unlist(d$values))
gender = as.integer(keys[, 4]),
year = as.integer(keys[, 5]),
n = as.integer(unlist(d$values))
)
}
batches <- split(munis, ceiling(seq_along(munis) / 50))
batches <- split(munis, ceiling(seq_along(munis) / 50))
raw_list <- vector("list", length(batches))
for (i in seq_along(batches)) {
cat(" Fetching batch", i, "/", length(batches),
"(munis", batches[[i]][1], "", tail(batches[[i]], 1), ")\n")
cat(
" Fetching batch",
i,
"/",
length(batches),
"(munis",
batches[[i]][1],
"",
tail(batches[[i]], 1),
")\n"
)
raw_list[[i]] <- tryCatch(
fetch_batch(batches[[i]]),
error = function(e) {
@ -93,7 +106,6 @@ for (i in seq_along(batches)) {
}
raw <- bind_rows(compact(raw_list))
cat("Total rows fetched:", nrow(raw), "\n")
# Aggregate: sum across ages and genders → n per (code, year, edu_level)
attainment <- raw |>
@ -104,14 +116,11 @@ attainment <- raw |>
attainment_summary <- attainment |>
group_by(code, year) |>
summarise(
n_total = sum(n),
n_total = sum(n),
n_postsec = sum(n[edu_level >= 5]),
pct_postsec = 100 * n_postsec / n_total,
.groups = "drop"
)
write_rds(attainment, "data/processed/attainment_ts.rds")
write_rds(attainment, "data/processed/attainment_ts.rds")
write_rds(attainment_summary, "data/processed/attainment_summary.rds")
cat("Saved attainment_ts.rds and attainment_summary.rds\n")
cat("Years:", paste(sort(unique(attainment$year)), collapse = ", "), "\n")
cat("Municipalities:", n_distinct(attainment$code), "\n")