Cfrgtkky

I am trying to get this command 'missing_compare' from the 'finalfit' package to work for my dataset:

proced<- c(1,NA,0,1,0,1,0)

asa<- c(4,3,4,2,5,1,NA)

albumin<- c(NA, NA, 3.572, NA, NA, NA, 4.262)

death<- c(0,0,1,0,1,1,0)

bmi<- c(26.04, NA, 31.23, 36.93, 28.9, NA, 30.01)

dataframe = data.frame(proced, as, albumin, death, bmi)

(This data frame is actually a lot bigger)

Then:

  dataframe$death = factor(dataframe$death)

  dataframe$proced = factor(dataframe$proced)

  dataframe$asa = factor(dataframe$asa)

And then:

explanatory = c("proced", "asa", 

                "bmi", "albumin")

dependent = "death"





dataframe %>% 

  summary_factorlist(dependent, explanatory, 

                     na_include=TRUE, p=TRUE)

But I can't get this to work:

 dataframe %>% 

  missing_compare(dependent, explanatory)

I get this error when I try to do the missing_compare command with my entire dataset:

Error in `[.default`(x, , 2) : subscript out of bounds

In addition: Warning messages:

1: In cor(x, rank(y)) : the standard deviation is zero

2: In cor(x, rank(y)) : the standard deviation is zero

Help!

As stated by @astrofunkswag, the purpose of this function is to compare the distribution of missingness across a particular variable.

You would be better starting with a visualisation of your missing data e.g.


dataframe %>%
missing_pairs(dependent, explanatory)


That will help you understand what data you have.

One sweats away at extensive vignettes and people accuse you of sparse documentation :)
http://finalfit.org/articles/missing.html

Let me know if you still can't get it to work.

Your dependent variable death has no missing values, which is the point of using the missing_compare function. Check out the documentation for that function for more info, though it is pretty sparse.

The missing_compare function compares an explanatory variable when the dependent variable is missing to the explanatory variable when the dependent variable is not missing. It applies tests to analyze whether these 2 are from the same distribution.

Using your example to illustrate this (note I reduce the number of DV's for simplicity):

explanatory = c("proced", "bmi")

dependent = "death"



dataframe2 <- dataframe

dataframe2$death[3:4] = NA



dataframe2 %>% 

  missing_compare(dependent, explanatory)



Missing data analysis: death           Not missing  Missing     p

2                       proced         0    2 (66.7) 1 (33.3) 1.000

3                                      1    2 (66.7) 1 (33.3)      

1                          bmi Mean (SD)    28.3 (2) 34.1 (4) 0.058

Warning message:

In chisq.test(tab, correct = FALSE) :

  Chi-squared approximation may be incorrect

I added 2 NA values to the dependent variable death, and the code runs. So for example the function is comparing the bmi values when death is missing to the values when it is not. The p column indicates whether the difference between the groups is statistically significant (Chi-Squared and Kruskal-Wallis for the continuous vars). I'd caution against relying solely on a p-value for this type of analysis, but that is irrelevant to how the code works.

And welcome to Stack Overflow!

Edit: great vignette