Loop to run regression on multiple dependent variables

Question

I have data such as this. I would like to run a linear regression on multiple variables separately without having to copy paste the regression.

dat <- read_table2("condition   school  Q2_1    Q2_2    Q2_4    Q2_8    Q2_10   Q2_11   Q2_14   Q2_15
1   A   3   4   3   2   2   4   4   2
0   B   3   3   3   2   1   2   2   1
1   C   4   4   4   3   3   4   3   3
0   D   3   4   3   3   2   2   4   2
1   A   2   4   2   3   3   3   3   3
0   B   2   4   4   2   3   2   2   3
1   C   3   3   3   2   3   3   2   3
0   D   4   4   3   3   3   2   2   3
1   A   3   3   3   3   2   3   3   2
0   B   3   3   3   2   3   3   4   1
1   C   4   4   4   4   3   3   3   3
0   D   3   3   4   3   3   4   4   2
1   A   2   2   2   2   2   2   2   3
0   B   3   3   3   2   2   2   2   3
1   C   3   4   3   1   2   3   3   3
0   D   3   4   2   2   3   4   3   3
1   A   4   4   4   3   3   4   3   2
0   B   4   2   3   2   3   2   2   1
1   C   4   3   3   4   3   4   3   3
0   D   3   3   3   2   2   2   3   2
1   A   4   2   3   3   2   2   2   3
")

I could run this separately and then pull out the coefficients.

 model1 <- lmer(Q2_1~ (1|school) + condition ,data=regression_data , REML = F)
 model2 <- lmer(Q2_2~ (1|school) + condition ,data=regression_data , REML = F)
 model3 <- lmer(Q2_4~ (1|school) + condition ,data=regression_data , REML = F)

summary(model1)$coefficient
summary(model2)$coefficient
summary(model3)$coefficient

But I rather do this in one chunk of code. I adjusted some code online to come up with this:

# outcome
out_start=3
out_end= 10
out_nvar=out_end-out_start+1

out_variable=rep(NA, out_nvar)
out_beta=rep(NA, out_nvar)
out_se = rep(NA, out_nvar)
out_pvalue=rep(NA, out_nvar)

# exposure
exp_start=1
exp_end=2
exp_nvar=exp_end-exp_start+1

exp_variable=rep(NA, exp_nvar)
exp_beta=rep(NA, exp_nvar)
exp_se = rep(NA, out_nvar)
exp_pvalue=rep(NA, exp_nvar)

number=1

library(lme4)
for (i in out_start:out_end){
  outcome = colnames(dat)[i]
  for (j in exp_start:exp_end){
    exposure = colnames(dat)[j]
    model <- lmer(get(outcome) ~ get(exposure) + (1|school) + condition, 
                  data=dat , REML = F, na.action = na.exclude)
    
    Vcov <- vcov(model, useScale = FALSE)
    beta <- fixef(model)
    se <- sqrt(diag(Vcov))
    zval <- beta / se
    pval <- 2 * pnorm(abs(zval), lower.tail = FALSE)
    
    out_beta[number] = as.numeric(beta[2])
    out_se[number] = as.numeric(se[2])
    out_pvalue[number] = as.numeric(pval[2])
    out_variable[number] = outcome
    number = number + 1
    
    exp_beta[number] = as.numeric(beta[2])
    exp_se[number] = as.numeric(se[2])
    exp_pvalue[number] = as.numeric(pval[2])
    exp_variable[number] = exposure
    number = number + 1
  }
}


outcome = data.frame(out_variable, out_beta, out_se, out_pvalue)
exposure = data.frame(exp_variable, exp_beta, exp_se, exp_pvalue)


library(tidyverse)
outcome = outcome %>% 
  dplyr::rename(
    variable = out_variable,
    beta = out_beta,
    se = out_se,
    pvalue = out_pvalue,
  )

exposure = exposure %>% 
  dplyr::rename(
    variable = exp_variable,
    beta = exp_beta,
    se = exp_se,
    pvalue = exp_pvalue,
  )
all = rbind(outcome, exposure)
all = na.omit(all)

This does not give me quite what I want -- it gives me the impact of being in the condition group for each dependent variable (question) but doe not show me the value of the intercept.

For example, for Q2_1 I expect to see:

model_lmer <- lmer(Q2_1~(1|school) + condition ,data=dat , REML = F)
summary(model_lmer)$coefficient
             Estimate Std. Error df    t value     Pr(>|t|)
(Intercept) 3.1000000  0.2079585 21 14.9068177 1.213221e-12
condition   0.1727273  0.2873360 21  0.6011334 5.541851e-01

Instead I see:

     var         beta        se           p value 
1   Q2_1     1.727273e-01   0.2873360   0.54775114
3   Q2_1    -2.559690e-15   0.3737413   1.00000000

Any suggestions are appreciated!

Duck · Accepted Answer

You can try this base R solution. I hope this helps:

library(lme4)
#Apply
models <- function(x)
{
  model1 <- lmer(x~ (1|school) + condition ,data=dat , REML = F)
  return(model1)
}


y <- apply(dat[,-c(1,2)],2,models)
#Extract results
extract <- function(x)
{
  z <- as.data.frame(summary(x)$coefficient)
  z$id <- rownames(z)
  z <- z[,c(dim(z)[2],1:(dim(z)[2]-1))]
  rownames(z)<-NULL
  return(z)
}
#Extract summary with function
DF <- as.data.frame(do.call(rbind,lapply(y,extract)))
#Format variables
DF$var <- gsub("\..*","",rownames(DF))
#Arrange columns
DF <- DF[,c(dim(DF)[2],1:(dim(DF)[2]-1))]
rownames(DF)<-NULL

     var          id     Estimate Std. Error     t value
1   Q2_1 (Intercept)  3.100000000  0.2079585 14.90681768
2   Q2_1   condition  0.172727273  0.2873360  0.60113340
3   Q2_2 (Intercept)  3.300000000  0.2251503 14.65687832
4   Q2_2   condition  0.063636364  0.3110898  0.20455947
5   Q2_4 (Intercept)  3.100000000  0.1928371 16.07574475
6   Q2_4   condition -0.009090909  0.2664427 -0.03411956
7   Q2_8 (Intercept)  2.300000000  0.2212714 10.39447415
8   Q2_8   condition  0.427272727  0.3057304  1.39754743
9  Q2_10 (Intercept)  2.500000000  0.1855144 13.47604443
10 Q2_10   condition  0.045454545  0.2563249  0.17733173
11 Q2_11 (Intercept)  2.500000000  0.2404001 10.39933014
12 Q2_11   condition  0.681818182  0.3321605  2.05267707
13 Q2_14 (Intercept)  2.800000000  0.2313147 12.10472106
14 Q2_14   condition  0.018181818  0.3196072  0.05688801
15 Q2_15 (Intercept)  2.100000000  0.2079585 10.09816681
16 Q2_15   condition  0.627272727  0.2873360  2.18306339

Loop to run regression on multiple dependent variables

Answers (1)

Related Questions