CODEWITHSUNDEEP
rdata.table
roody
roody

Reputation: 2663

Using data.table to create a column of regression coefficients

I'm struggling with what seems like it should be a simple extension of a previous question I'd asked here.

I'm trying to aggregate over (a) a range of dates and (b) a factor variable. Sample data might be:

Brand    Day     Rev     RVP              
  A      1        2535.00  195.00 
  B      1        1785.45  43.55 
  C      1        1730.87  32.66 
  A      2        920.00   230.00
  B      2        248.22   48.99 
  C      3        16466.00 189.00      
  A      1        2535.00  195.00 
  B      3        1785.45  43.55 
  C      3        1730.87  32.66 
  A      4        920.00   230.00
  B      5        248.22   48.99 
  C      4        16466.00 189.00

Thanks to helpful advice, I've figured out how to find the mean revenue for brands over a period of days using data.table:

new_df<-df[,(mean(Rev)), by=list(Brand,Day)]

Now, I'd like to create a new table where there is a column listing the coefficient estimate from an OLS regression of Rev by Day for each brand. I tried to do this as follows:

new_df2<-df[,(lm(Rev~Day)), by=list(Brand)]

That doesn't seem quite right. Thoughts? I'm sure it's something obvious I've missed.

Upvotes: 4

Views: 5467

Answers (3)

mnel
mnel

Reputation: 115392

You have a couple of options.

You could save the whole model object as a list within the data.table

models <- df[,   list(model  = list(lm(Rev ~ Day))),by = Brand]

models
   Brand model
1:     A  <lm>
2:     B  <lm>
3:     C  <lm>


# look at the models
models[,print(model[[1]]),by= Brand]


Call:
lm(formula = Rev ~ Day)

Coefficients:
(Intercept)          Day  
     2804.2       -538.3  



Call:
lm(formula = Rev ~ Day)

Coefficients:
(Intercept)          Day  
     1741.5       -263.5  



Call:
lm(formula = Rev ~ Day)

Coefficients:
(Intercept)          Day  
      -3698         4653

You can save the coefficients

models[, {coefs <- coef(model[[1]])
      list(coefs = coefs, name = names(coefs))}, by = Brand]
##     Brand      coefs        name
## 1:     A  2804.1667 (Intercept)
## 2:     A  -538.3333         Day
## 3:     B  1741.5291 (Intercept)
## 4:     B  -263.5251         Day
## 5:     C -3697.8621 (Intercept)
## 6:     C  4653.1989         Day

Or you can just extract the list of models

  models[,model]

Upvotes: 6

IRTFM
IRTFM

Reputation: 263342

> DF <- read.table(text="Brand    Day     Rev     RVP              
+   A      1        2535.00  195.00 
+   B      1        1785.45  43.55 
+   C      1        1730.87  32.66 
+   A      2        920.00   230.00
+   B      2        248.22   48.99 
+   C      3        16466.00 189.00      
+   A      1        2535.00  195.00 
+   B      3        1785.45  43.55 
+   C      3        1730.87  32.66 
+   A      4        920.00   230.00
+   B      5        248.22   48.99 
+   C      4        16466.00 189.00", header=TRUE)
> DT <- data.table(DF)
> Mod.tbl<-DT[, list(mod=list(lm(Rev~Day))), by=list(Brand)]
> Mod.tbl[ , coef(mod[[1]])["Day"], by= Brand]
   Brand           V1
1:     A -538.3333333
2:     B -263.5251429
3:     C 4653.1989474

Upvotes: 1

Marius
Marius

Reputation: 60070

I think this is what you want:

new_df2<-df[,(lm(Rev~Day)$coefficients[["Day"]]), by=list(Brand)]

lm returns a full model object, you need to drill down into it to get a single value from each group that can be turned into a column.

Upvotes: 3

Related Questions