I have a large dataset that I am trying to trim down for focus. As part of it I have three variables that are measured at five times. I've made each time into a mean so I ended up with five variables (one for each time) which I want to find the internal consistency of. When I try to calculate Cronbach's alpha using psych::alpha()
I get an error.
25 rows of reproducible data:
structure(list(catme_satis1a = c(4L, 4L, 5L, 5L, 5L, NA, 1L,
4L, 4L, 4L, 4L, 2L, 4L, 4L, 3L, 4L, 4L, 5L, 3L, 4L, 5L, 3L, 4L,
4L, 5L), catme_satis1b = c(4L, 4L, 4L, 5L, 5L, NA, 1L, 4L, 5L,
5L, 4L, 2L, 5L, 4L, 3L, 4L, 4L, 5L, 3L, 4L, 5L, 3L, 4L, 4L, 5L
), catme_satis1c = c(3L, 4L, 5L, 5L, 5L, NA, 1L, 4L, 3L, 4L,
4L, 2L, 4L, 5L, 3L, 4L, 4L, 5L, 3L, 4L, 5L, 3L, 4L, 4L, 5L),
catme_satis2a = c(4L, 4L, 4L, 5L, 5L, NA, 5L, 4L, 5L, NA,
NA, 3L, NA, 4L, 3L, 4L, 4L, 5L, 3L, NA, 5L, 5L, 4L, 4L, 5L
), catme_satis2b = c(4L, 4L, 5L, 5L, 5L, NA, 5L, 4L, 5L,
NA, NA, 3L, NA, 4L, 3L, 4L, 3L, 5L, 2L, NA, 5L, 5L, 4L, 4L,
5L), catme_satis2c = c(4L, 4L, 5L, 5L, 5L, NA, 5L, 4L, 5L,
NA, NA, 3L, NA, 4L, 3L, 4L, 3L, 5L, 3L, NA, 5L, 5L, 4L, 4L,
5L), catme_satis3a = c(4L, 4L, 4L, 5L, 5L, 5L, 4L, 4L, 5L,
5L, 3L, NA, 3L, 4L, 3L, NA, 4L, 5L, 3L, 5L, 5L, 5L, 4L, 5L,
5L), catme_satis3b = c(4L, 4L, 4L, 5L, 5L, 5L, 3L, 4L, 5L,
5L, 3L, NA, 3L, 4L, 3L, NA, 4L, 5L, 3L, 5L, 5L, 5L, 4L, 5L,
5L), catme_satis3c = c(4L, 5L, 5L, 5L, 5L, 4L, 4L, 4L, 5L,
5L, 3L, NA, 3L, 4L, 3L, NA, 4L, 5L, 4L, 5L, 5L, 5L, 4L, 4L,
5L), catme_satis4a = c(4L, 5L, 5L, 5L, 5L, 5L, 4L, 4L, 5L,
4L, 3L, 3L, 3L, 4L, 3L, 4L, 4L, 4L, 4L, NA, 5L, 5L, 5L, 5L,
5L), catme_satis4b = c(4L, 4L, 5L, 5L, 5L, 4L, 4L, 4L, 5L,
4L, 3L, 3L, 2L, 4L, 3L, 4L, 5L, 5L, 4L, NA, 5L, 5L, 5L, 5L,
5L), catme_satis4c = c(4L, 4L, 5L, 5L, 5L, 4L, 4L, 4L, 5L,
3L, 3L, 3L, 2L, 4L, 3L, 5L, 4L, 4L, 4L, NA, 5L, 5L, 5L, 5L,
5L), catme_satis5a = c(5L, 4L, 5L, 5L, 5L, 5L, 5L, 4L, 5L,
4L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 5L, 4L, 5L, 5L, 1L, 5L,
5L), catme_satis5b = c(5L, 4L, 5L, 5L, 5L, 5L, 5L, 4L, 5L,
4L, 3L, 3L, 3L, 4L, 4L, 5L, 4L, 5L, 5L, 4L, 5L, 5L, 1L, 5L,
5L), catme_satis5c = c(5L, 4L, 5L, 5L, 5L, 5L, 5L, 4L, 5L,
4L, 3L, 3L, 2L, 4L, 4L, 5L, 4L, 3L, 5L, 4L, 5L, 5L, 1L, 5L,
5L)), class = "data.frame", row.names = c(NA, -25L), .Names = c("catme_satis1a",
"catme_satis1b", "catme_satis1c", "catme_satis2a", "catme_satis2b",
"catme_satis2c", "catme_satis3a", "catme_satis3b", "catme_satis3c",
"catme_satis4a", "catme_satis4b", "catme_satis4c", "catme_satis5a",
"catme_satis5b", "catme_satis5c"))
Next, I tried this to get the mean for columns 1:3, 4:6, etc. (but by name):
library(dplyr)
df1 <- test %>%
rowwise() %>%
transmute(catme_satis1 = mean(c(catme_satis1a, catme_satis1b, catme_satis1c)),
catme_satis2 = mean(c(catme_satis2a, catme_satis2b, catme_satis2c)),
catme_satis3 = mean(c(catme_satis3a, catme_satis3b, catme_satis3c)),
catme_satis4 = mean(c(catme_satis4a, catme_satis4b, catme_satis4c)),
catme_satis5 = mean(c(catme_satis5a, catme_satis5b, catme_satis5c)))
Finally, I want to know the consistency of these variables using the psych
package:
library(psych)
alpha(df1)
Which gives this error:
> alpha(df1)
Error in sort.list(y) : 'x' must be atomic for 'sort.list'
Have you called 'sort' on a list?
My data frame seems to be correct when I print it, and I should be able to get the consistency of these values. Why is r
throwing this error?
After doing some exploration I found a way to make this work. It involves the dplyr
output having additional classes beyond data.frame
. I created the mean columns in a different manner to keep things away from dplyr
using the following code (note that this one is named df2
to facilitate comparison later on):
df2 <- data.frame(
catme_satis1 = apply(test[, 1:3], 1, mean),
catme_satis2 = apply(test[, 4:6], 1, mean),
catme_satis3 = apply(test[, 7:9], 1, mean),
catme_satis4 = apply(test[, 10:12], 1, mean),
catme_satis5 = apply(test[, 13:15], 1, mean)
)
The alpha(df2)
command worked just fine. This inspired me to check a few things about the dataframes.
The class of df1
from my original post, and df2
here is different:
> class(df1)
[1] "rowwise_df" "tbl_df" "tbl" "data.frame"
> class(df2)
[1] "data.frame"
Also, they recognized as being completely identical unless I coerce the dplyr output to be a dataframe!
> identical(df1, df2)
[1] FALSE
> identical(as.data.frame(df1), df2)
[1] TRUE
Running the command alpha(as.data.frame(df1))
works and produces identical results. There are two solutions here:
dplyr
methods to get the mean data. This keeps the data as a data.frame classed object.as.data.frame()
to coerce the object into the right class when running the alpha()
function. Or add %>% as.data.frame()
to the end of the dplyr
mutate command.