Home / Questions / 336621
Accepted
kev
Asked: 2020-03-13 08:45:53 +0800 CST 2020-03-13 08:45:53 +0800 CST

How to carry out regressions taking the dependent variable fixed and only the dependent variable changing?

  • 772

How to perform regressions taking the dependent variable fixed and changing only the dependent variable? That is, the data frame is something like

year  a b c d..... z
1
2
3
4

I need the dependent variable to always be abut the independent variable to be b, then cand so on and then give me the result in a list of the dependent variables that were significant, for example:

b
c
h

I tried to do the following

for (i in 1:40) {

  modelo_prueba <- lm(data = ds, formula = a ~ ds[ ,i]) 
  print(summary(modelo_prueba))

  }
r

1 Answers

  1. Best Answer

    Define the formula like this: it a ~ ds[ ,i]is syntactically valid, but the problem is that in the context of evaluating the linear model, neither the object inor the object exists ds.

    You can define the formula dynamically with each iteration as a simple string and then transform it into a real formula by as_formula():

    modelos <- list()
var_dep <- "a"
var_ind <- colnames(data)[colnames(data) != var_dep]
columnas_a_eliminar <- vector()
for (var in var_ind) {
  formula <- as.formula(paste(var_dep, "~ ", var))
  fit <- lm(data = data, formula = formula) 
  if (summary(fit)$coefficients[2,4] > 0.05)
    columnas_a_eliminar <- c(columnas_a_eliminar, var)
}

if (length(columnas_a_eliminar)>0)
  data <- data[, !(colnames(data) %in% columnas_a_eliminar),  drop = FALSE]

    In the example, we define what it is var_depand then the independent variables, taking all the columns except the one of the dependent variable, in your case, you would have (I imagine) remove the column yearas well. We then iterate through the names of each independent variable, generate a string for the formula, and transform it into such.

    • 1