Getting summaries from multiple simple linear regressions in R - r

Question

I am performing simple linear regressions on multiple groups from my data set. However, I want extract the summaries from each of these regressions and put them into a master table by organized by group. I can run it like this (and it works):
fit_basic <- rs2_anova %>% #Run multiple simple linear regressions
group_by(quant_method) %>%
nest() %>%
mutate(model = map(data, ~lm(recoveries ~ treatment, data = .)))
fit_basic_A <- fit_basic[[1,"model"]] #Remove the model from fit_basic
fit_basic_B <- fit_basic[[1,"model"]] #Remove the model from fit_basic
fit_basic_table_A <- get_regression_table(fit_basic_A) %>%
select("term", "estimate") %>%
pivot_wider(names_from = "term", values_from = "estimate") %>%
mutate(quant_method = "A")
fit_basic_table_B <- get_regression_table(fit_basic_A) %>%
select("term", "estimate") %>%
pivot_wider(names_from = "term", values_from = "estimate") %>%
mutate(quant_method = "B")
fit_basic_table <- rbind(fit_basic_table_A, fit_basic_table_B)
To save myself some lines of code (because I have many more groups than presented here) I thought I could use the map function, but I keep getting stuck at mapping the summary table, which throws an error:
fit_basic <- rs2_anova %>%
group_by(quant_method) %>%
nest() %>%
mutate(model = map(data, ~lm(recoveries ~ treatment, data = .))) %>%
mutate(summaries = map(data, get_regression_table(.$model)))
Error in input_checks(model, digits, print) :
Only simple linear regression models are supported. Try again using only `lm()` models as appropriate.
I also tried something along this line:
fit_basic_table <- map(fit_basic$model,
function(x) {
p <- get_regression_table(x)
cbind(par=rownames(p), p)
})
But I get a list of dataframes that I can't breakdown into a single dataframe and I have lost my group designations. I have tried:
fit_basic_table <- map(fit_basic$model,
function(x) {
p <- get_regression_table(x)
cbind(par=rownames(p), p)
}) %>%
map_df(as_tibble, .id = "id")
and
fit_basic_table <- map(fit_basic$model,
function(x) {
p <- get_regression_table(x)
cbind(par=rownames(p), p)
}) %>%
unnest(cols = "id")
Any thoughts on how to automate this?
*Random test dataframe:
quant_method <- c("A", "A","A","A","A","A","A","A","A","A","B","B","B","B","B","B","B","B","B","B")
treatment <- c("x","x","x","x","x","y","y","y","y","y","x","x","x","x","x","y","y","y","y","y")
recoveries <-c("88","86","87","82","85","76","65","55","72","71","98","96","97","92","99","66",
"55","55","62","61")
rs2_anova <- data.frame(quant_method, treatment, recoveries)

Answer 1

Here is one solution using tidyverse and broom packages. It is slightly different from the purrr method you were attempting, but I think the result shows the terms that you were interested in extracting from the lm object (i.e., term and estimate).
library(tidyverse)
library(broom)
#Added the stringsAsFactors argument = F to avoid an error in the lm model
rs2_anova <- data.frame(quant_method,
treatment,
recoveries,
stringsAsFactors = F)
fit_basic <- rs2_anova %>%
#Group by quant_method column
group_by(quant_method) %>%
#do the linear models by grouping var
do(model = lm(recoveries ~ treatment, data = .)) %>%
#tidy lm object and order it as tibble
tidy(model)

Answer 2

I found the answer here: unnest a list column after modeling after group_by in r
and modified it to:
fit_cec <- rs2_anova %>%
group_by(quant_method) %>%
nest %>%
mutate(data = map(data, ~ .x %>%
summarise(model = list(broom::tidy(lm(recoveries ~ treatment)))))) %>%
unnest(data) %>%
unnest(model)
or to get all estimates, predictions and summaries (https://drsimonj.svbtle.com/running-a-model-on-separate-groups). This also works nicely:
fit_cec <- rs2_anova %>%
group_by(quant_method) %>%
nest %>%
mutate(fit = map(data, ~ lm(loss_abs_BC_2 ~ cec, data = .)),
parameters = map(fit, tidy), #provides estimate table for slope and y-intercept with std.error and estimate p-values
summary = map(fit, broom::glance), #provides R2, adj.R2, sigma, and model p-value
predictions = map(fit, augment)) %>% #provides fitted values with residuals and errors
unnest(parameters) %>%
pivot_wider(names_from = term, values_from = c(estimate, std.error, statistic, p.value)) %>%
unnest(summary) %>%
unnest(predictions)