Advertisements
I need to merge the results of my regressions in a unique table. To give you an idea:
This is a sample of my dataset
> head(final, 20)
nquest nord sex anasc ireg eta staciv studio tpens
1 173 1 1 1948 18 64 3 5 2500
2 375 1 2 1925 16 87 4 2 1340
3 629 1 1 1939 5 73 4 3 1188
4 632 1 1 1950 5 62 1 3 1320
5 633 1 2 1934 5 78 4 2 350
6 1238 1 1 1937 15 75 4 3 1000
7 7886 1 1 1950 9 62 1 5 2000
8 11972 2 1 1938 17 74 1 2 750
9 20174 1 1 1941 8 71 1 5 2000
10 20174 2 2 1942 8 70 1 3 132
11 20223 1 2 1938 3 74 1 5 800
12 20223 2 1 1939 3 73 1 4 980
13 20711 2 1 1944 4 68 1 2 1900
14 20837 1 1 1931 8 81 1 4 1600
15 20837 2 2 1928 8 84 1 2 430
16 21461 1 2 1918 5 94 4 2 600
17 22173 1 1 1938 15 74 1 2 1200
18 22208 1 2 1935 5 77 4 2 700
19 22222 1 1 1927 5 85 4 2 1100
20 22276 1 1 1949 8 63 2 5 1170
and I’m running these kind of regressions
lm(log(tpens) ~ sex, data = final)
lm(log(tpens) ~ sex + eta, data = final)
lm(log(tpens) ~ sex + eta + ireg, data = final)
lm(log(tpens) ~ sex + eta + ireg + studio, data = final)
Is there a way to have the outputs next to each other??
This is an example of what I’m looking for (if it can include more information from the regressions, it would be better)
Estimate
(Intercept) 7.47635*** 8.5236948*** 8.5814025*** 7.4580630***
sex -0.42052*** -0.4170048*** -0.4229487*** -0.4153185***
eta -0.0146341*** -0.0145885*** -0.0068207***
ireg -0.0057238*** -0.0035033***
studio 0.1624156***
.... .....
I’ve seen from past questions that mapply and do.call can be used, but I’m not able to set them in the correct way… Can anyone help me? Is there another way to do it?
UPDATE
thanks to @akrun
# A tibble: 14 x 6
fmla term estimate std.error statistic p.value
<chr> <chr> <dbl> <dbl> <dbl> <dbl>
1 log(tpens) ~ sex (Intercept) 7.48 0.0221 339. 0
2 log(tpens) ~ sex sex -0.421 0.0148 -28.3 5.53e-162
3 log(tpens) ~ sex + eta (Intercept) 8.52 0.0648 132. 0
4 log(tpens) ~ sex + eta sex -0.417 0.0144 -29.0 2.92e-169
5 log(tpens) ~ sex + eta eta -0.0146 0.000854 -17.1 1.12e- 63
6 log(tpens) ~ sex + eta + ireg (Intercept) 8.58 0.0658 130. 0
7 log(tpens) ~ sex + eta + ireg sex -0.423 0.0144 -29.4 4.43e-173
8 log(tpens) ~ sex + eta + ireg eta -0.0146 0.000853 -17.1 1.44e- 63
9 log(tpens) ~ sex + eta + ireg ireg -0.00572 0.00125 -4.57 5.06e- 6
10 log(tpens) ~ sex + eta + ireg + studio (Intercept) 7.46 0.0602 124. 0
11 log(tpens) ~ sex + eta + ireg + studio sex -0.415 0.0119 -34.8 9.36e-234
12 log(tpens) ~ sex + eta + ireg + studio eta -0.00682 0.000729 -9.36 1.24e- 20
13 log(tpens) ~ sex + eta + ireg + studio ireg -0.00350 0.00104 -3.37 7.68e- 4
14 log(tpens) ~ sex + eta + ireg + studio studio 0.162 0.00367 44.2 0
UPDATE 2
# A tibble: 5 x 5
term `log(tpens) ~ sex` `log(tpens) ~ sex + eta` `log(tpens) ~ sex + eta + ireg` log(tpens) ~ sex + eta + ireg ~1
<chr> <dbl> <dbl> <dbl> <dbl>
1 (Intercept) 7.48 8.52 8.58 7.46
2 sex -0.421 -0.417 -0.423 -0.415
3 eta NA -0.0146 -0.0146 -0.00682
4 ireg NA NA -0.00572 -0.00350
5 studio NA NA NA 0.162
# ... with abbreviated variable name 1: `log(tpens) ~ sex + eta + ireg + studio`
>Solution :
We could loop over the formulas, create the linear model, extract the info needed (tidy) and return a single data
library(dplyr)
library(purrr)
library(broom)
out <- lst(log(tpens) ~ sex, log(tpens) ~ sex + eta,
log(tpens) ~ sex + eta + ireg,
log(tpens) ~ sex + eta + ireg + studio) %>%
map(lm, data = final) %>%
map(tidy) %>%
list_rbind(names_to = 'fmla')
If we want a wide format of ‘estimate’
library(tidyr)
out %>%
select(fmla, term, estimate) %>%
pivot_wider(names_from = fmla, values_from = estimate)