I am trying to run a Oaxaca decomposition using the oaxaca package, but the inclusion of certain variables seems to trigger the error "non-conformable arguments." As far as I can tell, the error seems to only arise with the inclusion of certain factor/categorical variables, but not all factor/categorical variables.
Here is a minimal reproducible example of my dataset, wvs_reduc:
structure(list(emp = c(1, 1, 1, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0,
1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0,
0, 0, 0, 0, 0, 0), education = structure(c(4L, 3L, 2L, 2L, 3L,
3L, 2L, 6L, 4L, 2L, 2L, 2L, 2L, 2L, 1L, 2L, 4L, 4L, 1L, 2L, 4L,
4L, 4L, 4L, 4L, 4L, 3L, 4L, 4L, 4L, 4L, 3L, 2L, 4L, 4L, 4L, 3L,
2L, 4L, 3L), .Label = c("No Formal Education", "Primary or Less",
"Incomplete Secondary", "Secondary", "Incomplete University",
"University or More"), class = "factor"), marital = structure(c(1L,
1L, 3L, 3L, 1L, 3L, 3L, 1L, 1L, 3L, 3L, 1L, 3L, 4L, 3L, 1L, 1L,
4L, 3L, 1L, 3L, 4L, 1L, 3L, 3L, 3L, 3L, 1L, 3L, 4L, 4L, 4L, 4L,
3L, 3L, 4L, 3L, 3L, 4L, 3L), .Label = c("single", "cohabiting",
"married", "previously married"), class = "factor"), Arab = c(1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)), row.names = c(NA,
-40L), class = c("tbl_df", "tbl", "data.frame"))
When I run the command:
library(oaxaca)
oaxaca(emp ~ education + marital | Arab,
data = wvs_reduc, group.weights = 0, R = 10)
I get the error message: Error in t(x.mean.A) %*% delta.A : non-conformable arguments.
In case it is relevant, when I run the command on my larger dataset, I instead get a similar but not-identical error with the inclusion of the variable "marital" but not "education" or other factor variables:
Error in t(x.mean.A - x.mean.B) %*% beta.B : non-conformable arguments
Looking at the underlying code oaxaca:::.oaxaca.wrap
and the error part is this bunch of lines:
E <- as.numeric(t(x.mean.A - x.mean.B) %*% beta.B)
C <- as.numeric(t(x.mean.B) %*% (beta.A - beta.B))
I <- as.numeric(t(x.mean.A - x.mean.B) %*% (beta.A - beta.B))
If anyone of x.mean.A is a vector, then it will throw an error. Looking at your design in this example dataset:
table(wvs_reduc$education,wvs_reduc$Arab)
0 1
No Formal Education 0 2
Primary or Less 2 10
Incomplete Secondary 4 3
Secondary 14 4
Incomplete University 0 0
University or More 0 1
So those will all zeros will be dropped and I would say you need to ensure the levels are distributed across your grouping category. We can affirm this by simulating this variable:
set.seed(111)
wvs_reduc$test_education =sample(levels(wvs_reduc$education),nrow(wvs_reduc),replace=TRUE)
wvs_reduc$test_marital =sample(levels(wvs_reduc$marital),nrow(wvs_reduc),replace=TRUE)
We run this and turn off bootstrap:
oaxaca(emp ~ test_education + test_marital | Arab, data=wvs_reduc,R=NULL)
And if we set bootstrap it crashes because when subsampling, it can run into the same error:
oaxaca(emp ~ test_education + test_marital | Arab, data=wvs_reduc,R=2)
oaxaca: oaxaca() performing analysis. Please wait.
Bootstrapping standard errors:
1 / 2 (50%)
Error in t(x.mean.A) %*% delta.A : non-conformable arguments
In addition: There were 11 warnings (use warnings() to see them)
So for it to work on your whole dataframe, you need to check whether there are levels with n=1 (considering groups)
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.