I'm an R beginner and keep banging my head over this. Is there any way to get the below problem solved?

I have a ts object (called mg.ts) with values for all of my x variables (gdp for 10 countries) and y variables (consumption for 10 countries), and there are no NAs in this dataset. I'm trying to do a for loop to do multiple regressions to loop through each country's gdp - consumption pair.

I made 2 vectors each with the names of all my x variables and y variables and tried to run a nested for loop to run the regression and grab out the corresponding coefficients, but keep getting an error.

I also need to eventually store these coefficients in a dataframe to analyze later, so if anyone knows how to do that easily that would be helpful too.

Sample data (mg.ts):

     X pakgdp belgdp pakcons belcons
2011 1 100    100    100     100     
2012 2 102    102    103     103    
2013 3 105    102    103     106    
2014 4 110    102    106     108    
2015 5 115    104    108     107
2016 6 120    105    111     108
2017 7 128    107    119     109
2018 8 133    109    125     111

My code:

x <- c("pakgdp", "belgdp", "usagdp",...)
y <- c("pakcons", "belcons", "usagdp",...)

for (i in x) {
  for (j in y) {
        as.formula(paste(i, “~”, j))
        dynlm( j ~ i , data=mg.ts)
        summary(dynlm( j ~ i , data=mg.ts))
  }
}

This is the error I'm getting:

Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) : 
  contrasts can be applied only to factors with 2 or more levels

In addition:

Warning message:
In storage.mode(v) <- "double" : NAs introduced by coercion

I'm an R beginner and keep banging my head over this. Is there any way to get the below problem solved?

I have a ts object (called mg.ts) with values for all of my x variables (gdp for 10 countries) and y variables (consumption for 10 countries), and there are no NAs in this dataset. I'm trying to do a for loop to do multiple regressions to loop through each country's gdp - consumption pair.

I made 2 vectors each with the names of all my x variables and y variables and tried to run a nested for loop to run the regression and grab out the corresponding coefficients, but keep getting an error.

I also need to eventually store these coefficients in a dataframe to analyze later, so if anyone knows how to do that easily that would be helpful too.

Sample data (mg.ts):

     X pakgdp belgdp pakcons belcons
2011 1 100    100    100     100     
2012 2 102    102    103     103    
2013 3 105    102    103     106    
2014 4 110    102    106     108    
2015 5 115    104    108     107
2016 6 120    105    111     108
2017 7 128    107    119     109
2018 8 133    109    125     111

My code:

x <- c("pakgdp", "belgdp", "usagdp",...)
y <- c("pakcons", "belcons", "usagdp",...)

for (i in x) {
  for (j in y) {
        as.formula(paste(i, “~”, j))
        dynlm( j ~ i , data=mg.ts)
        summary(dynlm( j ~ i , data=mg.ts))
  }
}

This is the error I'm getting:

Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) : 
  contrasts can be applied only to factors with 2 or more levels

In addition:

Warning message:
In storage.mode(v) <- "double" : NAs introduced by coercion

• r
• regression

• It's easier to help you if you include a simple reproducible example with sample input and desired output that can be used to test and verify possible solutions. Include all library() calls for the packages you use. Avoid using ... since that means we can't copy/paste the code for testing. Use dput() to share sample data. Use the formula you create inside your function: dynlm( as.formula(paste(i, “~”, j)) , data=mg.ts). Do you want i~j or j~i? – MrFlick Commented Mar 3 at 15:49

1 Answer 1

Your main problem is essentially a typo: you're creating a formula with as.formula(), but then you're not using it in your dynlm call — instead, you're trying to use j and i directly. As these are characters, not symbols, R doesn't know what to do with them.

However, I've written some code below that handles the two-way combinations of variables slightly more automatically and combines the results into a data frame.

library(dynlm)
library(broom)
library(purrr)
library(dplyr)

create sample data

dd <- read.table(header = TRUE, text = "
X pakgdp belgdp pakcons belcons
2011 1 100    100    100     100     
2012 2 102    102    103     103    
2013 3 105    102    103     106    
2014 4 110    102    106     108    
2015 5 115    104    108     107
2016 6 120    105    111     108
2017 7 128    107    119     109
2018 8 133    109    125     111
")
tsdat <- ts(dd[,-1], start = 2011)

fitting function

lmfun <- function(nm) {
    res <- dynlm(reformulate(nm[2], response = nm[1]), data = tsdat)
    ## force variable name evaluation
    res$call$formula <- eval(res$call$formula)
    res
}

fit all combinations

lmfitlist <- combn(colnames(tsdat), m = 2, lmfun, simplify = FALSE)

process fits/store as tibble

## set up data frame of response/predictor names
vars <- combn(colnames(tsdat), 2, rbind, simplify = FALSE) |>
    do.call(what = "rbind") |>
    as.data.frame() |>
    setNames(c("response", "predictor")) |>
    mutate(run = as.character(seq(n())), .before = 1)
purrr::map_dfr(lmfitlist, broom::tidy, .id = "run") |> 
    dplyr::full_join(x = vars, by = "run") |>
    as_tibble()

# A tibble: 12 × 8
   run   response predictor term        estimate std.error statistic     p.value
   <chr> <chr>    <chr>     <chr>          <dbl>     <dbl>     <dbl>       <dbl>
 1 1     pakgdp   belgdp    (Intercept) -297.      35.4        -8.40 0.000155   
 2 1     pakgdp   belgdp    belgdp         3.96     0.341      11.6  0.0000243  
 3 2     pakgdp   pakcons   (Intercept)  -36.5     12.6        -2.90 0.0275     
 4 2     pakgdp   pakcons   pakcons        1.38     0.115      12.0  0.0000204