For Loop for Regression based on variable name in R

Question

I'm currently running a series of regressions on a data set consisting of 72,000 observations with roughly 20 variables for each observation. One of these variables consists of 56 names and i want to run a regression for each name. I imagine I would create a for loop for this however I am a little inexperienced working with data sets of this size.
The variable containing the name is not in the regression. NAME : variable I want to run a for loop for to run regressions for each name.

My Code:

 my.mods = lapply(s.dat, FUN = function(x) {
 lm(log(TM+1000) ~ log(Inc+1) + log(Slip_sq_K+1) +
log(Ten+1) + log(HF+1) + log(BroP+1) + log(B+1) +
log(sian+1) + log(H_+1) + log(C_65+1) + log(D+1) +
log(TIP+1) + log(p+1) + log(X34itP+1) + log(FGK+1) +
log(X19tP+1) + log(X2nitP+1) + log(Car_AloneP+1) +
log(CaoledP+1) +log(PTsP+1) + log(Gy+1) +
log(OthemeansP+1) + log(HP+1) + log(Coi+1) +
log(electr) + log(Na+1),
data = x, na.action=na.exclude)

} )

Thanks!

Answer 1

I prefer to work without loops if possible and I found this page really helpful. It shows how you can use a model within a plyr function and get a table with all the important parameters back from it.

Answer 2

No need for loops. Just split your data and use lapply.

acs.dat.split = split(acs.dat, acs.dat$NAME)

my.mods = lapply(acs.dat, FUN = function(x) {
  lm(log(TSM+1000) ~ log(Inc+1) + log(Slip_sq_K+1) +
    log(Teen+1) + log(HFGG+1) + log(BrownP+1) + log(BlackP+1) +
    log(AsianP+1) + log(H_65+1) + log(C_65+1) + log(Detachedp+1) +
    log(TIP+1) + log(X2Unitp+1) + log(X34UnitP+1) + log(FGK+1) +
    log(X189tP+1) + log(X20PlusUnitP+1) + log(Car_AloneP+1) +
    log(CarpooledP+1) +log(PublicTransP+1) + log(Gly+1) +
    log(OthermeansP+1) + log(HomeP+1) + log(CommTime+1) +
    log(electr) + log(Natural_gas+1),
    data = x, na.action=na.exclude)
  } 
)

Answer 3

If you want to use loops you could subset the data for each name:

data(mtcars)
models = list()
for (i in 1:length(unique(row.names(mtcars)))) {
  sub_cars <- subset(x = mtcars, subset = row.names(mtcars) == row.names(mtcars)[i])
  models[i] <- lm(mpg ~ cyl, data = sub_cars)
}

Answer 4

Here is my solution in bare R, just a little bit long because I generated a code because I'am not sure if I get what you want. But I think if I get it you can just use the last line.

    # Random code for example
    set.seed(1)
    names <- c("Homer", "Bart", "Lisa")
    da <- rnorm(30)
    da1 <- rnorm(30, 2)
    data <- data.frame(Names = rep(names, 10), da, da1)

And here what I believe you can use:

reg <- by(data[, 2:3], data$Names, lm)

Here he output:

reg
data$Names: Bart

Call:
FUN(formula = data[x, , drop = FALSE])

Coefficients:
(Intercept)          da1  
     0.7738      -0.3076  

-------------------------------------------------------------- 
data$Names: Homer

Call:
FUN(formula = data[x, , drop = FALSE])

Coefficients:
(Intercept)          da1  
    0.14672     -0.01079  

-------------------------------------------------------------- 
data$Names: Lisa

Call:
FUN(formula = data[x, , drop = FALSE])

Coefficients:
(Intercept)          da1  
    -1.3974       0.7396