2

I want to run a panel regression on my data, consisting of 192 observations (48 observations each year, 4 years in total).

Since there are missing observations, I used "mice" package in R to impute my data (the imputation is done 10 times). After imputation, I used aggregation method to merge the 10 imputed dataset (Indeed, I used the exact method mentioned here)

Finally, I built the panel regression model with plm package. However, I saw the following message after the plm command.

Error in x[, !na.check] : incorrect number of dimensions

I don't understand why I have problems with the dimensions of my data. I can't find relevant resource here on stackoverflow as well. Readers can view my R codes below. 

dat <- read.csv("panel.csv")

# apply the mice package for imputation
library(mice)
impute <- mice(dat,m = 10, maxit = 50,meth = 'pmm',seed = 500)
dat2 <- complete(impute, action = "long")

### Aggregate dataset for model building
# Find the most frequent level in a factor variable
getmode <- function(v) {
  levels(v)[which.max(table(v))]
}

# Return either the mean or mode depending on the type of variable passed to it
my_summary <- function(x, id, ...){
if (is.numeric(x)) {
  return(tapply(x, id, mean))
}  
if (is.factor(x)) {
  return(tapply(x, id, getmode))
}  }

# Finally, use lapply to calculate the summaries
dat_panel <- data.frame(lapply(dat2, my_summary, id = dat2$.id))


### Now, we will built a panel model with fixed effects
# First, we build a model for ROA
library(plm)
reg <- plm(formula <- ROA ~ Year.of.IPO + TTassets + Debt.ratio + TTdonation,
        index <- c("Year","Level"),
        data <- dat_panel,
        method <- "within")

Can someone please explain why there is an error? I appreciate your explanations!

I am providing more information about my data. Indeed, I am using "dat_panel" to built my regression model. The dimension of the dataset is 192 rows x 11 columns. The structure of the dataset can be seen below

dim(dat_panel)
[1] 192  11

str(dat_panel)
'data.frame':   192 obs. of  11 variables:
 $ .imp       : chr [1:192(1d)] "1" "1" "1" "1" ...
  ..- attr(*, "dimnames")=List of 1
 .. ..$ : chr  "1" "10" "100" "101" ...
 $ .id        : chr [1:192(1d)] "1" "10" "100" "101" ...
 ..- attr(*, "dimnames")=List of 1
 .. ..$ : chr  "1" "10" "100" "101" ...

I also provide the first few rows of my dataset (as requested by someone)

> head(dat_panel)
   Year          Level Stock.Code Year.of.IPO    ROA     ROE TTassets Debt.ratio TTdonation
1  2013  Conglomerates          1          45   8.52  10.150 19.87659  0.1130966   17.30594
18 2013  Conglomerates         19          46   4.47   6.200 19.66903  0.1525779   17.57671
24 2013  Conglomerates         25          33   9.25  18.820 16.36257  0.2645771   12.31321
29 2013  Conglomerates         30          26  -4.30  -7.920 13.16387  0.0000000   12.45592
32 2013 Consumer Goods         33          10 -16.63 -18.023 12.43860  0.3243540   12.23134
35 2013 Consumer Goods         36          44  -2.56  -2.250 13.05192  0.0000000   11.08727
michigan8mileroad
  • 31
  • 3
  • It's hard to know what the problem is without seeing the data (so you could `dput` part of the dataset if you need more help), but in the code you gave us you seem to use `dat` as the data for the regression, whereas your imputed data is `dat2`. Is that the problem? If not, what is the output of `dim(dat2)`? – Oriol Mirosa Aug 15 '17 at 13:02
  • Hi Oriol, I have already edited my post. Please have a look! – michigan8mileroad Aug 16 '17 at 08:36
  • I found out the reason by myself. Inside the plm command, I need to use "=" instead of "<-" for all of my components. It runs perfectly well once I use "=" – michigan8mileroad Aug 16 '17 at 09:16

0 Answers0