R barplot + sapply shuffles bars (although it shouldn't)

Question

I need to plot means and standard errors for 4 variables grouped by 2 (left vs. right hand). Here are the data:

left_start_mydata = read.table(text="condition  force
right_small 1.80523635404968
                               right_small  2.6420765093878
                               right_small  -0.814658993753841
                               right_small  -2.60104096307957
                               right_small  -1.98589533137477
                               right_small  3.40251831946075
                               right_small  -0.320129242153803
                               right_small  -2.98033170716285
                               right_small  1.89317065279704
                               right_small  -3.84882524848594
                               right_small  -3.98968367934259
                               right_small  1.10427581334271
                               right_large  -1.75347355221301
                               right_large  0.791286271808679
                               right_large  -2.0073148173165
                               right_large  -5.03908061365724
                               right_large  -3.21618785397385
                               right_large  3.15958835997412
                               right_large  -0.728320450803572
                               right_large  -0.754841068944837
                               right_large  1.26489177600709
                               right_large  -1.25150854925629
                               right_large  2.91927950249639
                               right_large  0.343070062995591
                               left_small   2.76611178207954
                               left_small   1.98555350876524
                               left_small   1.90443573003935
                               left_small   0.939363367617274
                               left_small   1.47248738494375
                               left_small   -1.04761679029031
                               left_small   -0.824572467883381
                               left_small   -1.54423800803017
                               left_small   1.5187848305815
                               left_small   1.0956007263072
                               left_small   3.89244539291397
                               left_small   1.72801660622873
                               left_large   0.902501901614639
                               left_large   2.89567274148723
                               left_large   -0.503732000967399
                               left_large   -2.87429518370343
                               left_large   -1.85785327815289
                               left_large   -4.73265776308004
                               left_large   -0.752958593136438
                               left_large   2.47010977406911
                               left_large   -1.19149141260447
                               left_large   -0.396960252581726
                               left_large   1.54175722591051
                               left_large   2.05533917545533
                                ",header=TRUE)

At the next step, I calculate descriptive statistics for each condition:

attach(left_start_mydata)
left_start_mean_force = tapply(force, INDEX=condition, mean)   #means
left_start_sem_force = tapply(force,INDEX=condition,sd)/ sqrt(tapply(force,condition, length) ) #stand_errors

Now I plot:

barcols = c("red","blue")

sapply(2, 
       function(x) {
         mids = barplot(matrix(left_start_mean_force,
                               nrow=2,
                               byrow=TRUE),
                              ylim=c(-2,3),
                              beside=TRUE,
                              col=barcols)

         axis(1,at=colMeans(mids),
              c("left hand","right hand"),lwd=0,lwd.tick=0)

         abline(h=0)

         arrows(mids, left_start_mean_force - left_start_sem_force, 
                mids, left_start_mean_force + left_start_sem_force, 
                 code = 3, 
                 angle = 90, 
                 length = 0.1, 
                 lwd = 2)

       }
)

And I get almost what I need (see the figure below).

BUT! If you look at the bars for the right hand, you can see that the red one (which should represent condition "right_large") is actually lower than the blue bar next to it ("right_small"), whereas the actual value is higher (i.e., closer to zero):

> left_start_mean_force
 left_large  left_small right_large right_small 
  1.2812381  -0.6430682  -0.5242566  -0.6063786 

It seems that somehow these two columns were shuffled. The problem only appears for mean values. Standard errors are represented correctly, i.e., "right_large" is shown on the left, and "right_small" on the right side.

What is the problem? I assume, it must be something either with the barplot function of with sapply.

P.S.: please don't suggest me ggplot and other packages, I'm sure there's a simple soluton with standard functions.

Answer 1

Your data is grouped incorrectly. You need to set byrow to FALSE in this part of the code:

     mids = barplot(matrix(left_start_mean_force,
                           nrow=2,
                           byrow=FALSE), # <<<<<< HERE
                          ylim=c(-2,3),
                          beside=TRUE,
                          col=barcols)

When you pass a matrix to barplot, it groups values by column, not row.

 m = matrix(1:4, nrow=2)
 barplot(m, beside=T) 
 # m is:
 #   1  3
 #   2  4

On the other hand, what you are doing is similar to this:

 m = matrix(1:4, nrow=2, byrow=T)
 barplot(m, beside=T) 
 # m is:
 #   1  2
 #   3  4