使用 dplyr 拟合多个回归模型 [英] Fitting several regression models with dplyr

问题描述

我想使用 dplyr 为每小时(因子变量)拟合一个模型，但出现错误，我不太确定是什么问题.

I would like to fit a model for each hour(the factor variable) using dplyr, I'm getting an error, and i'm not quite sure what's wrong.

df.h <- data.frame( 
  hour     = factor(rep(1:24, each = 21)),
  price    = runif(504, min = -10, max = 125),
  wind     = runif(504, min = 0, max = 2500),
  temp     = runif(504, min = - 10, max = 25)  
)

df.h <- tbl_df(df.h)
df.h <- group_by(df.h, hour)

group_size(df.h) # checks out, 21 obs. for each factor variable

# different attempts:
reg.models <- do(df.h, formula = price ~ wind + temp)

reg.models <- do(df.h, .f = lm(price ~ wind + temp, data = df.h))

我尝试了各种变体，但我无法让它发挥作用.

I've tried various variations, but I can't get it to work.

推荐答案

截至 2020 年中期，tchakravarty's回答 会失败.为了规避 broom 和 dpylr 似乎交互的新方法，下面的 broom::tidy, broom 组合::augment 和 broom::glance 都可以使用.我们只需要在 do() 和稍后的 unnest() tibble 中使用它们.

As of mid 2020, tchakravarty's answer will fail. In order to circumvent the new approach of broom and dpylr seem to interact, the following combination of broom::tidy, broom::augment and broom::glance can be used. We just have to use them inside do() and later unnest() the tibble.

library(dplyr)
library(broom)

df.h = data.frame( 
  hour     = factor(rep(1:24, each = 21)),
  price    = runif(504, min = -10, max = 125),
  wind     = runif(504, min = 0, max = 2500),
  temp     = runif(504, min = - 10, max = 25)  
)

df.h %>% group_by(hour) %>%
  do(fitHour = tidy(lm(price ~ wind + temp, data = .))) %>% 
  unnest(fitHour)
# # A tibble: 72 x 6
#    hour  term        estimate std.error statistic   p.value
#    <fct> <chr>          <dbl>     <dbl>     <dbl>     <dbl>
#  1 1     (Intercept)   82.4     18.1         4.55  0.000248 
#  2 1     wind         -0.0212   0.0108      -1.96  0.0655   
#  3 1     temp         -1.01     0.792       -1.28  0.218    
#  4 2     (Intercept)   25.9     19.7         1.31  0.206    
#  5 2     wind          0.0204   0.0131       1.57  0.135    
#  6 2     temp          0.680    1.01         0.670 0.511    
#  7 3     (Intercept)   88.3     15.5         5.69  0.0000214
#  8 3     wind         -0.0188   0.00998     -1.89  0.0754   
#  9 3     temp         -0.669    0.653       -1.02  0.319    
# 10 4     (Intercept)   73.4     14.2         5.17  0.0000639

df.h %>% group_by(hour) %>%
  do(fitHour = augment(lm(price ~ wind + temp, data = .))) %>% 
  unnest(fitHour)
# # A tibble: 24 x 13
#    hour  r.squared adj.r.squared sigma statistic p.value    df logLik   AIC   BIC deviance
#    <fct>     <dbl>         <dbl> <dbl>     <dbl>   <dbl> <dbl>  <dbl> <dbl> <dbl>    <dbl>
#  1 1        0.246        0.162    39.0     2.93   0.0790     2  -105.  218.  222.   27334.
#  2 2        0.161        0.0674   43.5     1.72   0.207      2  -107.  223.  227.   34029.
#  3 3        0.192        0.102    33.9     2.14   0.147      2  -102.  212.  217.   20739.
#  4 4        0.0960      -0.00445  34.3     0.956  0.403      2  -102.  213.  217.   21169.
#  5 5        0.230        0.144    31.7     2.68   0.0955     2  -101.  210.  214.   18088.
#  6 6        0.0190      -0.0900   39.8     0.174  0.842      2  -106.  219.  223.   28507.
#  7 7        0.0129      -0.0967   37.1     0.118  0.889      2  -104.  216.  220.   24801.
#  8 8        0.197        0.108    35.3     2.21   0.139      2  -103.  214.  218.   22438.
#  9 9        0.0429      -0.0634   39.4     0.403  0.674      2  -105.  219.  223.   27918.
# 10 10       0.0943      -0.00633  35.6     0.937  0.410      2  -103.  214.  219.   22854.
# # … with 14 more rows, and 2 more variables: df.residual <int>, nobs <int>

df.h %>% group_by(hour) %>%
  do(fitHour = glance(lm(price ~ wind + temp, data = .))) %>% 
  unnest(fitHour)
# # A tibble: 504 x 10
#    hour   price  wind   temp .fitted .resid .std.resid   .hat .sigma  .cooksd
#    <fct>  <dbl> <dbl>  <dbl>   <dbl>  <dbl>      <dbl>  <dbl>  <dbl>    <dbl>
#  1 1      94.2   883. -6.64     70.4  23.7       0.652 0.129    39.6 0.0209  
#  2 1      19.3  2107.  2.40     35.4 -16.0      -0.431 0.0864   39.9 0.00584 
#  3 1      60.5  2161. 18.3      18.1  42.5       1.18  0.146    38.5 0.0795  
#  4 1     116.   1244. 12.0      44.0  71.9       1.91  0.0690   35.8 0.0902  
#  5 1     117.   1624. -8.05     56.1  60.6       1.67  0.128    36.9 0.136   
#  6 1      75.0   220. -0.838    78.6  -3.58     -0.101 0.175    40.1 0.000724
#  7 1     106.    765.  6.15     60.0  45.7       1.22  0.0845   38.4 0.0461  
#  8 1      -9.89 2055. 12.3      26.5 -36.4      -0.979 0.0909   39.0 0.0319  
#  9 1      96.1   215. -8.36     86.3   9.82      0.287 0.232    40.0 0.00830 
# 10 1      27.2   323. 22.4      52.9 -25.7      -0.777 0.278    39.4 0.0774  
# # … with 494 more rows

感谢 Bob Muenchen 的博客 获得灵感.

这篇关于使用 dplyr 拟合多个回归模型的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！