How to calculate slope of linear regression for 1000 data.frames?-CodePudding

df <- data.frame(a=rnorm(20, 0,1), b=rnorm(20,1,2), c=rnorm(20, 2, 1), d=rnorm(20, 1,2))
 a          b          c           d
1  -0.114975143  2.8923335  2.9968141  1.68054858
2  -0.354557896 -2.3895332  3.7244438  1.99219906
3   0.271021912 -1.7725649  1.3278507  2.54157377
4  -1.981331306  2.7159506  3.4150731  3.57905491
5   0.003403129 -2.4912678  0.7665065  0.78206113
6  -0.843816087  0.4654837  2.2260497  5.16645779
7   0.091269485  2.6912003  2.1541783 -0.25013719
8  -0.221281275  1.6365565 -0.3745516  0.48183139
9   0.041891592 -1.1115767  1.2828104  0.07650962
10  1.182289018 -1.3741108  0.5180835  2.28457132
11 -0.498770473  1.3988605  1.3275124 -2.86340984
12 -0.277915549 -2.1960097  3.8331209  5.24206680
13 -0.522938649  0.9404911  3.0822281 -0.10335791
14  0.318142859  0.8651728  1.9794013  1.06888933
15 -0.870920037  3.9335164  4.5851187  0.37296019
16 -0.536834653 -0.2343890  1.5840454  0.83641016
17  1.369788371 -1.0971564  1.1922038 -0.91191447
18  0.355511037  1.1537255  1.2379696 -0.31673585
19  0.279638498  0.2543774  2.6568435  0.37737600
20  0.515248897 -4.3723985  2.5923414  2.60326350

I have 1000 data frames like this. I would like to apply linear regression (say for eg: y = column b, and x = column d) for each data frame separately to find out slopes. I would like to get a list of slopes and if possible regression equations for all 1000 models.

CodePudding user response：

Create the formula with reformulate, loop over the list of data.frames, create the model with lm, extract the slope from the coefficients (coef)

form1 <- reformulate('d', response = 'b')
sapply(lst1, function(x) coef(lm(form1, data = x))[[2]])