Pairwise comparisons across columns from two different data frames

Question

Here is a toy dataframe, mydf

    M1  M2  M3  M4  M5  M6  M7  M8                      
A1  2   2   1   1   1   1   0   2
A2  0   1   2   2   1   1   0   2
A3  2   0   0   1   1   0   1   1
A4  2   0   1   2   2   1   1   0
A5  1   2   0   2   2   1   0   1
A6  2   1   1   2   2   1   1   1

And here is a screenshot of second dataframe, mismatches:

        M1      M2      M3      M4      M5  M6  M7      M8                              
A1-A2   TRUE    NA      NA      NA      NA  NA  FALSE   FALSE
A1-A3   FALSE   TRUE    NA      NA      NA  NA  NA      NA
A1-A4   FALSE   TRUE    NA      NA      NA  NA  NA      TRUE
A1-A5   NA      FALSE   NA      NA      NA  NA  FALSE   NA
A1-A6   FALSE   NA      NA      NA      NA  NA  NA      NA
A2-A1   TRUE    NA      NA      NA      NA  NA  FALSE   FALSE
A2-A3   TRUE    NA      TRUE    NA      NA  NA  NA      NA
A2-A4   TRUE    NA      NA      FALSE   NA  NA  NA      TRUE
A2-A5   NA      NA      TRUE    FALSE   NA  NA  FALSE   NA
A2-A6   TRUE    NA      NA      FALSE   NA  NA  NA      NA

Now I'd like to define a function to compare the two dataframes where mismatches[a,] == TRUE & mydf[b,] != 1). So for the trio A1-A2/A3, the value for column M1 should be TRUE since mismatches for A1-A2 == TRUE & mydf for A3 !=1. Here is a snapshot of my desired result

            M1      M2      M3  M4  M5  M6  M7      M8
A1-A2/A3    TRUE    NA      NA  NA  NA  NA  FALSE   FALSE
A1-A2/A4    TRUE    NA      NA  NA  NA  NA  FALSE   FALSE
A1-A2/A5    FALSE   NA      NA  NA  NA  NA  FALSE   FALSE
A1-A2/A6    TRUE    NA      NA  NA  NA  NA  FALSE   FALSE
A1-A3/A2    FALSE   FALSE   NA  NA  NA  NA  NA      NA
A1-A3/A4    FALSE   TRUE    NA  NA  NA  NA  NA      NA
A1-A3/A5    FALSE   TRUE    NA  NA  NA  NA  NA      NA
A1-A3/A6    FALSE   FALSE   NA  NA  NA  NA  NA      NA

I can't seem to find the question that I used as a model for my working code but here are some similar postsCount Number of Pairwise Differences...;Compare all the columns pairwise...;Pairwise ... ignoring empty values. Mismatches was derived from mydf by identifying 0/2 and 2/0 pairwise comparisons.

#make a 2/0 matrix for for calculating incompatibilities
nohets <- mydf
nohets[nohets == 1] <- NA
#the incompatibility function
incompatible <- function(a,b,x)
    x[a,]!=x[b,]

The idea was to use this same strategy to compare mydf and mismatches.

#make a trio function
trio_id <- function(a,b,x,y)
    x[a,] == TRUE & y[b,] !=1
rows <- c(row.names(mismatches),row.names(mydf))
grid <- subset(expand.grid(Var1=rows,Var2=rows))
impossible <- t(mapply(trio_id,grid$Var2,grid$Var1,
    MoreArgs=list(x=mismatches,y=mydf)))
dimnames(impossible) <- list(paste(grid$Var2,grid$Var1,
    sep="/"),names(mydf))

This runs, but does not give the desired result but rather

            M1      M2  M3  M4  M5  M6  M7      M8
A1-A2/A3    NA      NA  NA  NA  NA  NA  FALSE   FALSE
A1-A2/A4    NA      NA  NA  NA  NA  NA  FALSE   FALSE
A1-A2/A5    NA      NA  NA  NA  NA  NA  FALSE   FALSE
A1-A2/A6    NA      NA  NA  NA  NA  NA  FALSE   FALSE
A1-A3/A2    FALSE   NA  NA  NA  NA  NA  NA      NA
A1-A3/A4    FALSE   NA  NA  NA  NA  NA  NA      NA
A1-A3/A5    FALSE   NA  NA  NA  NA  NA  NA      NA
A1-A3/A6    FALSE   NA  NA  NA  NA  NA  NA      NA

I don't need every comparison calculated... and I suspect writing the code for every possible comparison in 'grid' is the source of my problem.

score 0 · Accepted Answer · answered Jan 23 '18 at 20:35

Does this work?

mydf<-matrix(sample(c(0,1,2),6*8,replace = T),ncol=8)
mismatches<-matrix(sample(c(T,F,NA),6*8,replace = T),ncol=8)
mydf<-mydf != 1

> (mydf+mismatches) == 2
      [,1]  [,2]  [,3]  [,4]  [,5]  [,6]  [,7]  [,8]
[1,]    NA    NA FALSE FALSE    NA FALSE  TRUE FALSE
[2,] FALSE    NA FALSE  TRUE  TRUE    NA FALSE FALSE
[3,]  TRUE    NA    NA FALSE FALSE  TRUE FALSE FALSE
[4,] FALSE FALSE FALSE    NA    NA    NA FALSE    NA
[5,]    NA FALSE FALSE    NA    NA FALSE  TRUE  TRUE
[6,] FALSE FALSE  TRUE FALSE    NA  TRUE FALSE  TRUE

Yes, this works, thanks! It looks like the matrices need to have the same dimensions for this to work. I just need to make sure I construct them correctly ahead of time. — Aaron Dickey, Jan 24 '18 at 14:18

Pairwise comparisons across columns from two different data frames

1 Answers1