Split, apply linear model, combine

Question

I have a dataframe as follows:

df

Date       Hour ID Par1  Par2  Par3 
08-01-15     0   A   2    3      4
08-01-15     0   B   4    5      6 
08-01-15     1   N   2    9      10
08-01-15     1   A   3    7      23
08-01-15     1   B   4    7      22
08-02-15     0   E   2    4      12
08-02-15     0   A   3    7       9

So I want to split this dataframe by Hour as follows:

splitdata<-split(df<-split(df, df$Hour)

After splitting it, I want to apply a linear model to the split dataset.

result <- lapply(splitdata, function(df){
  lm1 <-lm(Par1~Par2,data=df)
  summary <- (lm1$summary)
  data.frame(as.list(summary))
})
result

My results do not show anything. Though if I changed summary to:

summary <- (lm1$coef)
data.frame(as.list(summary))

Then it will produce a result.

So the main question is, how do I get a list of the summaries of each of the linear models by Hour and not just the coefficients?

Thanks!

@akrun a list of the summaries by hour, not just the coefficients. The one you just posted was great except it didn't have the summaries because I want to see the significance of Par2 — Nick, Aug 05 '15 at 16:40
You can't fit the entire summary in a data.frame - what specific numbers do you want?? — Señor O, Aug 05 '15 at 16:41
@SeñorO The Pr(>|t|) value in the coefficients table. I am trying to determine the significance of Par2 for each of the hours. — Nick, Aug 05 '15 at 16:45

score 6 · Accepted Answer · answered Aug 05 '15 at 16:42

6

Try using lmList from nlme

library(nlme)
fits <- lmList(Par1 ~ Par2 | Hour, data=df)

This splits the data by hour and fits linear models for you. Then, you can just do summary(fits). Or, you can look at each summary individually with

lapply(fits, summary)

answered Aug 05 '15 at 16:42

Rorschach

31,301
5
78
129

That worked perfectly! Thanks – Nick Aug 05 '15 at 16:47

score 3 · Answer 2 · answered Aug 05 '15 at 16:52

If you're looking for a table form (which I usually prefer):

> dt = data.table(df)
> dt[,{S = summary(lm(Par1 ~ Par2))$coefficients; list(Coef = S[2,1], Intercept = S[1,1], Sig = S[2,4])}, by = Hour]
   Hour      Coef Intercept      Sig
1:    0  0.314286   1.25714 0.439388
2:    1 -0.750000   8.75000 0.333333

Split, apply linear model, combine

2 Answers2