for 循环未在 R 中迭代

Question

I have a large df that follows the structure below. For each of 200 groups, I want to fit a linear model to 30 years of data, then extract the slope and the R squared.

   count <- c(5, 10, 15, 20, 2, 4, 6, 8, 1, 2, 3, 4, 10, 20, 30, 40)
   group <- c(0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3)
   year <- c(2000, 2001, 2002, 2003, 2000, 2001, 2002, 2003, 2000, 2001, 2002, 2003, 2000, 2001, 2002, 2003)
   df <- data.frame(count, group, year)

I am able to get the data I want for one group at a time like this

   group_1 <- filter(df, group == 1)
   lm_group1 <-  lm(count ~ year, data = group_1)
   lm_coef_1 <- lm_group_1$coefficients["year"]
   lm_rsq_1 <-  summary(lm_group_1)$adj.r.squared
   df_group_1 <- data.frame(lm_coef_1, lm_rsq_1)

Doing this 200 times via copy-paste is not sensible, so I am trying to automate with a for loop.

    for (i in df$group) {

         group_data <- filter(df, group == i)

         lm_g <-  lm(count ~ year, data = group_data) 
           lm_coef <- lm_g$coefficients["year"] 
           lm_rsq <- summary(lm_g)$adj.r.squared  
           df_i <- data.frame(lm_coef, lm_rsq)

         if (i == 199)
         break
    }

This runs and raises no errors. However, instead of producing 200 dfs, one for each group, it produces one single df, df_i.

I have tried naming all the variables with i (lm_i, lm_coef_i, lm_rsq_i, df_i); have tried inserting %>% between statements; have tried looking for examples of similar problems. Have tried to apply the tidy function from the broom package, but it seems to drop the r squared, or at least I can't find where it puts it.

Primarily, I want to know why this loop refuses to iterate. However, I am open to and appreciate other suggestions for how to solve this problem more elegantly.

Answer 1

Just putting i after a variable name won't automatically create a new object. I think what you want to do is make a list with 200 indices containing the individual data frames that you want.

#make a list the size of your expected output

my_df_list <- vector("list", 200)

 for (i in df$group) {

         group_data <- filter(df, group == i)

         lm_g <-  lm(count ~ year, data = group_data) 
           lm_coef <- lm_g$coefficients["year"] 
           lm_rsq <- summary(lm_g)$adj.r.squared  
           my_df_list[[i]] <- data.frame(lm_coef, lm_rsq)

         if (i == 199)
         break
    }

This way you can keep each of your data.frames separate while iterating through.

Answer 2

You could also do:

library(tidyverse)

df %>%
  group_by(group) %>%
  summarise(d = list(tibble(coef_year = coef(a<-lm(count~year, cur_data()))['year'],
                          adjusted_r_sq = summary(a)$adj.r.sq)))

Answer 3

We can use broom

library(dplyr)
library(broom)
library(tidyr)
df %>% 
  group_by(group) %>%
  summarise(out = list(glance(lm(count ~ year, data = cur_data())) %>% 
               select(r.squared, adj.r.squared))) %>%
 unnest(out)