group_by() in list of data frames

Question

I want to:

1. merge list out with dataframe df
2. estimate an lm() model

id <- c(1,2,3,4,5,1,2,3,4,5)
quarter <- c("1","2","1","1","2", "3","1","1","3","3")
month <- c(3,4,2,1,5,7,3,1,8,9)
pred_dif <- c(0.5,0.1,0.15,0.23,0.75,0.6,0.49,0.81,0.37,0.14)

list_1 <- data.frame(id, pred_dif, month)

pred_dif <- c(0.45,0.18,0.35,0.63,0.25,0.63,0.29,0.11,0.17,0.24)

list_2 <- data.frame(id, pred_dif, month)

pred_dif <- c(0.58,0.13,0.55,0.13,0.76,0.3,0.29,0.81,0.27,0.04)

list_3 <- data.frame(id, pred_dif, month)

pred_dif <- c(0.3,0.61,0.18,0.29,0.85,0.76,0.56,0.91,0.48,0.91)

list_4 <- data.frame(id, pred_dif, month)

out <- list(list_1, list_2, list_3, list_4)


pred_second <- c(0.4,0.71,0.28,0.39,0.95,0.86,0.66,0.81,0.58,0.81)
df <- data.frame(id, quarter, pred_second, month)



library(purrr)
library(dplyr)
library(broom)
library(tidyr)
lmout_lst <- map(out, 
                 ~ left_join(.x, df, by = c('id', 'month')) %>%
                   group_by(quarter) %>%
                   summarise(new = list(lm(pred_dif ~ as.factor(month) - 1) %>% 
                                          broom::tidy(.))) %>%
                   unnest(new))

The problem happens in ols_list_reg. In particular with the "group_by" command.

Any idea why this is happening and possible solutions?

Answer 1

Perhaps, you can try this -

library(tidyverse)

map(out, 
    ~ left_join(.x, df, by = c('id', 'month')) %>%
      group_by(quarter) %>%
      summarise(new = list({
            tryCatch(lm(pred_dif ~ as.factor(month) - 1) %>% broom::tidy(.), 
                     error = function(e) tibble(estimate = NA))
        })) %>%
      unnest(new)
)

If you want to combine all the results together use map_df instead of map.

Answer 2

As @RonakShah says, your code fails for an individual element of the list. It's not at all clear what you're trying to achieve, but

out %>% 
  bind_rows(.id="element") %>% 
  left_join(df, by=c("id", "period")) %>% 
  mutate(period=as.factor(period)) %>% 
  group_by(element) %>% 
  group_map(function(.x, .y) lm(pred_dif ~ period-1, data=.x))

at least runs without warning or error and gives possibly sensible output:

[[1]]

Call:
lm(formula = pred_dif ~ period - 1, data = .x)

Coefficients:
period01  period02  period08  period09  period11  period12  
   0.365     0.600     0.620     0.100     0.370     0.412  


[[2]]

Call:
lm(formula = pred_dif ~ period - 1, data = .x)

Coefficients:
period01  period02  period08  period09  period11  period12  
   0.540     0.630     0.270     0.180     0.170     0.232  


[[3]]

Call:
lm(formula = pred_dif ~ period - 1, data = .x)

Coefficients:
period01  period02  period08  period09  period11  period12  
   0.355     0.300     0.525     0.130     0.270     0.552  


[[4]]

Call:
lm(formula = pred_dif ~ period - 1, data = .x)

Coefficients:
period01  period02  period08  period09  period11  period12  
   0.295     0.760     0.705     0.610     0.480     0.618