R项目：lm给出错误后循环中断

Question

I want to automatically run linear regressions and save the results. The beta coefficients generated by R project will be later used as parameters for further computations.

This an example of my data layout:

 id |x_2000|x_2001|x_2002|y_2000|y_2001|y_2002|z_2000|z_2001|z_2002
 1  |20   |NA     |6     |90    |NA    |80    |54    |NA    |10
 2  |50   |NA     |10    |50    |NA    |50    |60    |NA    |40
 3  |4    |NA     |1     |5     |NA    |10    |30    |NA    |120

x is value x and the number behind it represents a year. The same logic applies to the other variables y and z.

To run the linear regressions, I have created a loop. I use the following code to loop through the variables and to run a regression for each year.

for (i in 2000:2002){
  X_COLUMN <- c(paste0("x_",i))
  Y_COLUMN <- c(paste0("y_",i))
  Z_COLUMN <- c(paste0("z_",i))
  result.lm <- lm(as.formula(paste("formula=",X_COLUMN,"~",Y_COLUMN,"+",Z_COLUMN,"-1")), data=data_for_regression)
  b1 <- rbind(b1, c(x,i,coef(result.lm)[1]))
  b2 <- rbind(b2, c(x,i,coef(result.lm)[2]))
  }

For 2000 everything works well, however when the loop continues to year 2001 it hits the NA values. This results in a error message:

Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, …) : 0 (non-NA) cases

Logical, because there is no complete case for year 2001. The result is that the loop breaks. However, I don't want it to break but to continue to the next year.

Any idea how to fix this?

Answer 1

How about using a try statement

for (i in 2000:2002){
    X_COLUMN <- c(paste0("x_",i))
    Y_COLUMN <- c(paste0("y_",i))
    Z_COLUMN <- c(paste0("z_",i))
    try({
        result.lm <- lm(as.formula(paste("formula=",X_COLUMN,"~",Y_COLUMN,"+",Z_COLUMN,"-1")), data=data_for_regression)
        b1 <- rbind(b1, c(x,i,coef(result.lm)[1]))
        b2 <- rbind(b2, c(x,i,coef(result.lm)[2]))
    }, silent=T)
}

or an tryCatch

for (i in 2000:2002){
    X_COLUMN <- c(paste0("x_",i))
    Y_COLUMN <- c(paste0("y_",i))
    Z_COLUMN <- c(paste0("z_",i))
    tryCatch({
        result.lm <- lm(as.formula(paste("formula=",X_COLUMN,"~",Y_COLUMN,"+",Z_COLUMN,"-1")), data=data_for_regression)
        b1 <- rbind(b1, c(x,i,coef(result.lm)[1]))
        b2 <- rbind(b2, c(x,i,coef(result.lm)[2]))
    }, error=function(e) {
        b1 <- rbind(b1, c(x,i,NA))
        b2 <- rbind(b2, c(x,i,NA))
    })
}