CODEWITHSUNDEEP
rregressionlm
Harmzy15
Harmzy15

Reputation: 497

Export summary of multiple regressions from list

I have a list of multiple regressions completed via this code using the standard dataset mtcars. 

models <- lapply(paste("mpg", names(mtcars)[-1], sep = "~"), formula)
res.models <- lapply(models, FUN = function(x) {summary(lm(formula = x, data = mtcars))})
names(res.models) <- paste("mpg", names(mtcars)[-1], sep = "~")

Where I now have a list of multiple regressions against the first column "mpg". From here I am trying to export certain summary statistics such as; intercept, coefficient and r.squared.

I have tried using a loop which I've included below.

for (i in 1:length(res.models))
{
  res <- res.models[[i]]
  res_bound <- NULL
  intercept <- res$coefficients[1]
  coef <- res$coefficients[2]
  r <- res$r.squared
  res_bound <- cbind(intercept, coef, r)
}

Although this gets me a dataframe it only includes the results from the last regression model, a 1 row by 3 column dataframe. Furthermore, I would like to have the "terms" of each regression in the table to distinguish between which model I am looking at (e.g. mpg vs cyl or mpg vs hp).

Am I simply missing a step in my loop? The ultimate goal is to write.csv the final dataframe.

Upvotes: 1

Views: 2824

Answers (2)

ulfelder
ulfelder

Reputation: 5335

If you want to do it in base R:

res <- lapply(seq_along(res.models), function(i) {

  data.frame(model = names(res.models)[i],
             intercept = res.models[[i]]$coefficients[1],
             coef = res.models[[i]]$coefficients[2],
             r = res.models[[i]]$r.squared,
             stringsAsFactors = FALSE)

})

do.call(rbind, res)

Output:

      model intercept        coef         r
1   mpg~cyl 37.884576 -2.87579014 0.7261800
2  mpg~disp 29.599855 -0.04121512 0.7183433
3    mpg~hp 30.098861 -0.06822828 0.6024373
4  mpg~drat -7.524618  7.67823260 0.4639952
5    mpg~wt 37.285126 -5.34447157 0.7528328
6  mpg~qsec -5.114038  1.41212484 0.1752963
7    mpg~vs 16.616667  7.94047619 0.4409477
8    mpg~am 17.147368  7.24493927 0.3597989
9  mpg~gear  5.623333  3.92333333 0.2306734
10 mpg~carb 25.872334 -2.05571870 0.3035184

The reason for seq_along(res.models) instead of just res.models is so we can also grab the name for the associated slot in the list and drop it into the data frame you're making.

Upvotes: 3

alistaire
alistaire

Reputation: 43334

You can use purrr::map_df to apply broom::glance to each model and then collect the results into a data.frame:

purrr::map_df(res.models, broom::glance, .id = 'formula')
#>     formula r.squared adj.r.squared    sigma statistic      p.value df
#> 1   mpg~cyl 0.7261800     0.7170527 3.205902 79.561028 6.112687e-10  2
#> 2  mpg~disp 0.7183433     0.7089548 3.251454 76.512660 9.380327e-10  2
#> 3    mpg~hp 0.6024373     0.5891853 3.862962 45.459803 1.787835e-07  2
#> 4  mpg~drat 0.4639952     0.4461283 4.485409 25.969645 1.776240e-05  2
#> 5    mpg~wt 0.7528328     0.7445939 3.045882 91.375325 1.293959e-10  2
#> 6  mpg~qsec 0.1752963     0.1478062 5.563738  6.376702 1.708199e-02  2
#> 7    mpg~vs 0.4409477     0.4223126 4.580827 23.662241 3.415937e-05  2
#> 8    mpg~am 0.3597989     0.3384589 4.902029 16.860279 2.850207e-04  2
#> 9  mpg~gear 0.2306734     0.2050292 5.373695  8.995144 5.400948e-03  2
#> 10 mpg~carb 0.3035184     0.2803024 5.112961 13.073646 1.084446e-03  2

You could do something similar with broom::tidy for the coefficients, or broom::augment for the residuals. Note that broom functions are intended to be called on the models themselves, not the summaries, but you can keep the whole thing in a pipeline, if you like:

library(purrr)

names(mtcars)[-1] %>% 
    paste('mpg ~', .) %>%    # or start with `models` at this point
    map(lm, data = mtcars) %>% 
    map_df(broom::glance, .id = 'formula')
#>    formula r.squared adj.r.squared    sigma statistic      p.value df
#> 1        1 0.7261800     0.7170527 3.205902 79.561028 6.112687e-10  2
#> 2        2 0.7183433     0.7089548 3.251454 76.512660 9.380327e-10  2
#> 3        3 0.6024373     0.5891853 3.862962 45.459803 1.787835e-07  2
#> 4        4 0.4639952     0.4461283 4.485409 25.969645 1.776240e-05  2
#> 5        5 0.7528328     0.7445939 3.045882 91.375325 1.293959e-10  2
#> 6        6 0.1752963     0.1478062 5.563738  6.376702 1.708199e-02  2
#> 7        7 0.4409477     0.4223126 4.580827 23.662241 3.415937e-05  2
#> 8        8 0.3597989     0.3384589 4.902029 16.860279 2.850207e-04  2
#> 9        9 0.2306734     0.2050292 5.373695  8.995144 5.400948e-03  2
#> 10      10 0.3035184     0.2803024 5.112961 13.073646 1.084446e-03  2
#>       logLik      AIC      BIC deviance df.residual
#> 1  -81.65321 169.3064 173.7036 308.3342          30
#> 2  -82.10469 170.2094 174.6066 317.1587          30
#> 3  -87.61931 181.2386 185.6358 447.6743          30
#> 4  -92.39996 190.7999 195.1971 603.5667          30
#> 5  -80.01471 166.0294 170.4266 278.3219          30
#> 6  -99.29406 204.5881 208.9853 928.6553          30
#> 7  -93.07356 192.1471 196.5443 629.5193          30
#> 8  -95.24219 196.4844 200.8816 720.8966          30
#> 9  -98.18192 202.3638 206.7611 866.2980          30
#> 10 -96.59033 199.1807 203.5779 784.2711          30

Note you get a few extra variables that can't aren't contained in the summary.

Upvotes: 5

Related Questions