benchmarks/04-glm-vs-eglm.R
In pachamaltese/boostedglm: Efficient Fitting of Linear and Generalized Linear Models
library(ggplot2)
library(dplyr)
library(bench)
library(ggplot2)
library(stringr)
load("benchmarks/01-chapter1-traditional-gravity-bench.RData")
bench_traditional_gravity <- bind_rows(
bench_fe %>% mutate(model = "TG, FE", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols %>% mutate(model = "TG, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols_remoteness %>% mutate(model = "TG, OLS REM", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "TG, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity", ls(), value = TRUE, invert = TRUE))
load("benchmarks/02-chapter1-distance-puzzle-bench.RData")
bench_distance_puzzle <- bind_rows(
bench_fe %>% mutate(model = "DP, FE", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols %>% mutate(model = "DP, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_home_bias %>% mutate(model = "DP, HB", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_internal_distance %>% mutate(model = "DP, ID", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "DP, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity|bench_distance_puzzle", ls(), value = TRUE, invert = TRUE))
load("benchmarks/03-chapter1-regional-trade-agreements-bench.RData")
bench_rtas <- bind_rows(
bench_ols %>% mutate(model = "RTAs, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_endogeneity %>% mutate(model = "RTAs, EN", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_globalization %>% mutate(model = "RTAs, GB", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_phasing %>% mutate(model = "RTAs, PH", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "RTAs, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_reverse_causality %>% mutate(model = "RTAs, RC", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_trade_diversion %>% mutate(model = "RTAs, TD", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity|bench_distance_puzzle|bench_rtas", ls(), value = TRUE, invert = TRUE))
bench_models <- bind_rows(
bench_traditional_gravity,
bench_distance_puzzle,
bench_rtas
)
bench_models <- bench_models %>%
# filter(fun %in% c("eglm", "glm")) %>%
mutate(`Function` = toupper(fun)) %>%
mutate(
median = str_trim(as.character(median)),
median_units = str_extract(median, "[a-z]+"),
median = as.numeric(str_remove(median, median_units)),
median_seconds = case_when(
median_units == "ms" ~ median / 1000,
median_units == "s" ~ median,
median_units == "m" ~ median * 60
),
mem_alloc = str_trim(as.character(mem_alloc)),
mem_alloc_units = str_extract(mem_alloc, "[A-Z]+"),
mem_alloc = as.numeric(str_remove(mem_alloc, mem_alloc_units)),
mem_alloc_mb = case_when(
mem_alloc_units == "MB" ~ mem_alloc,
mem_alloc_units == "GB" ~ mem_alloc * 1024
)
)
# colors <- c("#1c26b3", "#d05555", "#549e95", "#993f7b")
g2_time <- ggplot(bench_models %>% filter(fun %in% c("eglm", "glm"))) +
geom_col(aes(x = model, y = median_seconds, fill = `Function`), position = "dodge2") +
scale_fill_manual(values = c("#3B9AB2", "#E1AF00")) +
theme_minimal() +
# theme(legend.position = "none") +
labs(x = "Model", y = "Median Time (in Seconds)",
title = "GLM vs EGLM - Time Benchmark")
g2_memory <- ggplot(bench_models %>% filter(fun %in% c("eglm", "glm"))) +
geom_col(aes(x = model, y = mem_alloc_mb, fill = `Function`), position = "dodge2") +
scale_fill_manual(values = c("#3B9AB2", "#E1AF00")) +
theme_minimal() +
# theme(legend.position = "none") +
labs(x = "Model", y = "Allocated Memory (in MB)",
title = "GLM vs EGLM - Memory Usage Benchmark")
load("benchmarks/01-chapter1-traditional-gravity.RData")
object_size_traditional_gravity <- tibble(
model = c(rep("TG, OLS", 2), rep("TG, OLS REM", 2), rep("TG, FE", 2), rep("TG, PPML", 2)),
fun = c(rep(c("lm", "elm"), 3), c("glm", "eglm")),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_ols_remoteness_lm),
object.size(fit_ols_remoteness_elm),
object.size(fit_fe_lm),
object.size(fit_fe_elm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_traditional_gravity <- tibble(
model = c(rep("TG, OLS", 2), rep("TG, OLS REM", 2), rep("TG, FE", 2), rep("TG, PPML", 2)),
fun = c(rep(c("lm", "elm"), 3), c("glm", "eglm")),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_ols_remoteness_lm$qr$qr)[1],
fit_ols_remoteness_elm$qr$original.dimensions[1],
dim(fit_fe_lm$qr$qr)[1],
fit_fe_elm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_ols_remoteness_lm$qr$qr)[2],
fit_ols_remoteness_elm$qr$original.dimensions[2],
dim(fit_fe_lm$qr$qr)[2],
fit_fe_elm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2]
)
)
rm(fit_fe_elm, fit_fe_lm, fit_ols_elm, fit_ols_lm, fit_ols_remoteness_elm,
fit_ols_remoteness_lm, fit_ppml_eglm, fit_ppml_glm)
load("benchmarks/02-chapter1-distance-puzzle.RData")
object_size_distance_puzzle <- tibble(
model = c(rep("DP, OLS", 2), rep("DP, FE", 2), rep("DP, HB", 2), rep("DP, ID", 2), rep("DP, PPML", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 4)),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_fe_glm),
object.size(fit_fe_eglm),
object.size(fit_home_bias_glm),
object.size(fit_home_bias_eglm),
object.size(fit_internal_distance_glm),
object.size(fit_internal_distance_eglm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_distance_puzzle <- tibble(
model = c(rep("DP, OLS", 2), rep("DP, FE", 2), rep("DP, HB", 2), rep("DP, ID", 2), rep("DP, PPML", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 4)),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_fe_glm$qr$qr)[1],
fit_fe_eglm$qr$original.dimensions[1],
dim(fit_home_bias_glm$qr$qr)[1],
fit_home_bias_eglm$qr$original.dimensions[1],
dim(fit_internal_distance_glm$qr$qr)[1],
fit_internal_distance_eglm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_fe_glm$qr$qr)[2],
fit_fe_eglm$qr$original.dimensions[2],
dim(fit_home_bias_glm$qr$qr)[2],
fit_home_bias_eglm$qr$original.dimensions[2],
dim(fit_internal_distance_glm$qr$qr)[2],
fit_internal_distance_eglm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2]
)
)
rm(fit_fe_eglm, fit_fe_glm, fit_home_bias_eglm, fit_home_bias_glm,
fit_internal_distance_eglm, fit_internal_distance_glm, fit_ols_elm,
fit_ols_lm, fit_ppml_eglm, fit_ppml_glm)
load("benchmarks/03-chapter1-regional-trade-agreements.RData")
object_size_rtas <- tibble(
model = c(rep("RTAs, OLS", 2), rep("RTAs, END", 2), rep("RTAs, GB", 2), rep("RTAs, PHA", 2),
rep("RTAs, PPML", 2), rep("RTAs, REV", 2), rep("RTAs, DIV", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 6)),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_endogeneity_glm),
object.size(fit_endogeneity_eglm),
object.size(fit_globalization_glm),
object.size(fit_globalization_eglm),
object.size(fit_phasing_glm),
object.size(fit_phasing_eglm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm),
object.size(fit_reverse_causality_glm),
object.size(fit_reverse_causality_eglm),
object.size(fit_trade_diversion_glm),
object.size(fit_trade_diversion_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_rtas <- tibble(
model = c(rep("RTAs, OLS", 2), rep("RTAs, END", 2), rep("RTAs, GB", 2), rep("RTAs, PHA", 2),
rep("RTAs, PPML", 2), rep("RTAs, REV", 2), rep("RTAs, DIV", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 6)),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_endogeneity_glm$qr$qr)[1],
fit_endogeneity_eglm$qr$original.dimensions[1],
dim(fit_globalization_glm$qr$qr)[1],
fit_globalization_eglm$qr$original.dimensions[1],
dim(fit_phasing_glm$qr$qr)[1],
fit_phasing_eglm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1],
dim(fit_reverse_causality_glm$qr$qr)[1],
fit_reverse_causality_eglm$qr$original.dimensions[1],
dim(fit_trade_diversion_glm$qr$qr)[1],
fit_trade_diversion_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_endogeneity_glm$qr$qr)[2],
fit_endogeneity_eglm$qr$original.dimensions[2],
dim(fit_globalization_glm$qr$qr)[2],
fit_globalization_eglm$qr$original.dimensions[2],
dim(fit_phasing_glm$qr$qr)[2],
fit_phasing_eglm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2],
dim(fit_reverse_causality_glm$qr$qr)[2],
fit_reverse_causality_eglm$qr$original.dimensions[2],
dim(fit_trade_diversion_glm$qr$qr)[2],
fit_trade_diversion_eglm$qr$original.dimensions[2]
)
)
rm(fit_endogeneity_eglm, fit_endogeneity_glm, fit_globalization_eglm,
fit_globalization_glm, fit_ols_elm, fit_ols_lm, fit_phasing_eglm,
fit_phasing_glm, fit_ppml_eglm, fit_ppml_glm, fit_reverse_causality_eglm,
fit_reverse_causality_glm, fit_trade_diversion_eglm, fit_trade_diversion_glm)
save.image("benchmarks/04-glm-vs-eglm.RData")
pachamaltese/boostedglm documentation built on July 24, 2024, 10:30 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
library(ggplot2)
library(dplyr)
library(bench)
library(ggplot2)
library(stringr)
load("benchmarks/01-chapter1-traditional-gravity-bench.RData")
bench_traditional_gravity <- bind_rows(
bench_fe %>% mutate(model = "TG, FE", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols %>% mutate(model = "TG, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols_remoteness %>% mutate(model = "TG, OLS REM", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "TG, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity", ls(), value = TRUE, invert = TRUE))
load("benchmarks/02-chapter1-distance-puzzle-bench.RData")
bench_distance_puzzle <- bind_rows(
bench_fe %>% mutate(model = "DP, FE", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ols %>% mutate(model = "DP, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_home_bias %>% mutate(model = "DP, HB", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_internal_distance %>% mutate(model = "DP, ID", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "DP, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity|bench_distance_puzzle", ls(), value = TRUE, invert = TRUE))
load("benchmarks/03-chapter1-regional-trade-agreements-bench.RData")
bench_rtas <- bind_rows(
bench_ols %>% mutate(model = "RTAs, OLS", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_endogeneity %>% mutate(model = "RTAs, EN", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_globalization %>% mutate(model = "RTAs, GB", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_phasing %>% mutate(model = "RTAs, PH", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_ppml %>% mutate(model = "RTAs, PPML", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_reverse_causality %>% mutate(model = "RTAs, RC", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc),
bench_trade_diversion %>% mutate(model = "RTAs, TD", fun = gsub("\\(.*", "", expression)) %>% select(model, fun, median, mem_alloc)
)
rm(list = grep("bench_traditional_gravity|bench_distance_puzzle|bench_rtas", ls(), value = TRUE, invert = TRUE))
bench_models <- bind_rows(
bench_traditional_gravity,
bench_distance_puzzle,
bench_rtas
)
bench_models <- bench_models %>%
# filter(fun %in% c("eglm", "glm")) %>%
mutate(`Function` = toupper(fun)) %>%
mutate(
median = str_trim(as.character(median)),
median_units = str_extract(median, "[a-z]+"),
median = as.numeric(str_remove(median, median_units)),
median_seconds = case_when(
median_units == "ms" ~ median / 1000,
median_units == "s" ~ median,
median_units == "m" ~ median * 60
),
mem_alloc = str_trim(as.character(mem_alloc)),
mem_alloc_units = str_extract(mem_alloc, "[A-Z]+"),
mem_alloc = as.numeric(str_remove(mem_alloc, mem_alloc_units)),
mem_alloc_mb = case_when(
mem_alloc_units == "MB" ~ mem_alloc,
mem_alloc_units == "GB" ~ mem_alloc * 1024
)
)
# colors <- c("#1c26b3", "#d05555", "#549e95", "#993f7b")
g2_time <- ggplot(bench_models %>% filter(fun %in% c("eglm", "glm"))) +
geom_col(aes(x = model, y = median_seconds, fill = `Function`), position = "dodge2") +
scale_fill_manual(values = c("#3B9AB2", "#E1AF00")) +
theme_minimal() +
# theme(legend.position = "none") +
labs(x = "Model", y = "Median Time (in Seconds)",
title = "GLM vs EGLM - Time Benchmark")
g2_memory <- ggplot(bench_models %>% filter(fun %in% c("eglm", "glm"))) +
geom_col(aes(x = model, y = mem_alloc_mb, fill = `Function`), position = "dodge2") +
scale_fill_manual(values = c("#3B9AB2", "#E1AF00")) +
theme_minimal() +
# theme(legend.position = "none") +
labs(x = "Model", y = "Allocated Memory (in MB)",
title = "GLM vs EGLM - Memory Usage Benchmark")
load("benchmarks/01-chapter1-traditional-gravity.RData")
object_size_traditional_gravity <- tibble(
model = c(rep("TG, OLS", 2), rep("TG, OLS REM", 2), rep("TG, FE", 2), rep("TG, PPML", 2)),
fun = c(rep(c("lm", "elm"), 3), c("glm", "eglm")),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_ols_remoteness_lm),
object.size(fit_ols_remoteness_elm),
object.size(fit_fe_lm),
object.size(fit_fe_elm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_traditional_gravity <- tibble(
model = c(rep("TG, OLS", 2), rep("TG, OLS REM", 2), rep("TG, FE", 2), rep("TG, PPML", 2)),
fun = c(rep(c("lm", "elm"), 3), c("glm", "eglm")),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_ols_remoteness_lm$qr$qr)[1],
fit_ols_remoteness_elm$qr$original.dimensions[1],
dim(fit_fe_lm$qr$qr)[1],
fit_fe_elm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_ols_remoteness_lm$qr$qr)[2],
fit_ols_remoteness_elm$qr$original.dimensions[2],
dim(fit_fe_lm$qr$qr)[2],
fit_fe_elm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2]
)
)
rm(fit_fe_elm, fit_fe_lm, fit_ols_elm, fit_ols_lm, fit_ols_remoteness_elm,
fit_ols_remoteness_lm, fit_ppml_eglm, fit_ppml_glm)
load("benchmarks/02-chapter1-distance-puzzle.RData")
object_size_distance_puzzle <- tibble(
model = c(rep("DP, OLS", 2), rep("DP, FE", 2), rep("DP, HB", 2), rep("DP, ID", 2), rep("DP, PPML", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 4)),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_fe_glm),
object.size(fit_fe_eglm),
object.size(fit_home_bias_glm),
object.size(fit_home_bias_eglm),
object.size(fit_internal_distance_glm),
object.size(fit_internal_distance_eglm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_distance_puzzle <- tibble(
model = c(rep("DP, OLS", 2), rep("DP, FE", 2), rep("DP, HB", 2), rep("DP, ID", 2), rep("DP, PPML", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 4)),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_fe_glm$qr$qr)[1],
fit_fe_eglm$qr$original.dimensions[1],
dim(fit_home_bias_glm$qr$qr)[1],
fit_home_bias_eglm$qr$original.dimensions[1],
dim(fit_internal_distance_glm$qr$qr)[1],
fit_internal_distance_eglm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_fe_glm$qr$qr)[2],
fit_fe_eglm$qr$original.dimensions[2],
dim(fit_home_bias_glm$qr$qr)[2],
fit_home_bias_eglm$qr$original.dimensions[2],
dim(fit_internal_distance_glm$qr$qr)[2],
fit_internal_distance_eglm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2]
)
)
rm(fit_fe_eglm, fit_fe_glm, fit_home_bias_eglm, fit_home_bias_glm,
fit_internal_distance_eglm, fit_internal_distance_glm, fit_ols_elm,
fit_ols_lm, fit_ppml_eglm, fit_ppml_glm)
load("benchmarks/03-chapter1-regional-trade-agreements.RData")
object_size_rtas <- tibble(
model = c(rep("RTAs, OLS", 2), rep("RTAs, END", 2), rep("RTAs, GB", 2), rep("RTAs, PHA", 2),
rep("RTAs, PPML", 2), rep("RTAs, REV", 2), rep("RTAs, DIV", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 6)),
size = c(
object.size(fit_ols_lm),
object.size(fit_ols_elm),
object.size(fit_endogeneity_glm),
object.size(fit_endogeneity_eglm),
object.size(fit_globalization_glm),
object.size(fit_globalization_eglm),
object.size(fit_phasing_glm),
object.size(fit_phasing_eglm),
object.size(fit_ppml_glm),
object.size(fit_ppml_eglm),
object.size(fit_reverse_causality_glm),
object.size(fit_reverse_causality_eglm),
object.size(fit_trade_diversion_glm),
object.size(fit_trade_diversion_eglm)
)
) %>%
mutate(size = size / 1000^2)
model_matrix_rtas <- tibble(
model = c(rep("RTAs, OLS", 2), rep("RTAs, END", 2), rep("RTAs, GB", 2), rep("RTAs, PHA", 2),
rep("RTAs, PPML", 2), rep("RTAs, REV", 2), rep("RTAs, DIV", 2)),
fun = c("lm", "elm", rep(c("glm", "eglm"), 6)),
model_matrix_nrow = c(
dim(fit_ols_lm$qr$qr)[1],
fit_ols_elm$qr$original.dimensions[1],
dim(fit_endogeneity_glm$qr$qr)[1],
fit_endogeneity_eglm$qr$original.dimensions[1],
dim(fit_globalization_glm$qr$qr)[1],
fit_globalization_eglm$qr$original.dimensions[1],
dim(fit_phasing_glm$qr$qr)[1],
fit_phasing_eglm$qr$original.dimensions[1],
dim(fit_ppml_glm$qr$qr)[1],
fit_ppml_eglm$qr$original.dimensions[1],
dim(fit_reverse_causality_glm$qr$qr)[1],
fit_reverse_causality_eglm$qr$original.dimensions[1],
dim(fit_trade_diversion_glm$qr$qr)[1],
fit_trade_diversion_eglm$qr$original.dimensions[1]
),
model_matrix_ncol = c(
dim(fit_ols_lm$qr$qr)[2],
fit_ols_elm$qr$original.dimensions[2],
dim(fit_endogeneity_glm$qr$qr)[2],
fit_endogeneity_eglm$qr$original.dimensions[2],
dim(fit_globalization_glm$qr$qr)[2],
fit_globalization_eglm$qr$original.dimensions[2],
dim(fit_phasing_glm$qr$qr)[2],
fit_phasing_eglm$qr$original.dimensions[2],
dim(fit_ppml_glm$qr$qr)[2],
fit_ppml_eglm$qr$original.dimensions[2],
dim(fit_reverse_causality_glm$qr$qr)[2],
fit_reverse_causality_eglm$qr$original.dimensions[2],
dim(fit_trade_diversion_glm$qr$qr)[2],
fit_trade_diversion_eglm$qr$original.dimensions[2]
)
)
rm(fit_endogeneity_eglm, fit_endogeneity_glm, fit_globalization_eglm,
fit_globalization_glm, fit_ols_elm, fit_ols_lm, fit_phasing_eglm,
fit_phasing_glm, fit_ppml_eglm, fit_ppml_glm, fit_reverse_causality_eglm,
fit_reverse_causality_glm, fit_trade_diversion_eglm, fit_trade_diversion_glm)
save.image("benchmarks/04-glm-vs-eglm.RData")
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.