How to adjust heights of subplots in rmarkdown

I want to plot 6 scatterplots as subplots in one figure. I tried different ways of adjusting the height of the subplots but I do not find a satisfying solution.
Either the plots are very narrow so information can not be seen properly or - if I try to adjust height with absoulte values - I always get the error, my figure region is too large, even if I chose very small values.

I am working in RMarkdown and output to PDF

What I tried so far:

1. using layout() with relative widths and heights:

nf <- layout(matrix(c(1,2,3,4,5,6), ncol=2,nrow=3, TRUE),widths = c(1,1), heights=c(1,1,1))

plot(totDistM,OAM,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,OAW,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,PA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,PA_W,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,UA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,UA_W,pch = 21,col="black", bg="yellow", cex=0.7)

resulting in too narrow subplots:

2. using layout() with absolute heights:

nf <- layout(matrix(c(1,2,3,4,5,6), ncol=2,nrow=3, TRUE),widths = c(1,1), heights=c(lcm(4),lcm(4),lcm(4)))

plot(totDistM,OAM,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,OAW,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,PA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,PA_W,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,UA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,UA_W,pch = 21,col="black", bg="yellow", cex=0.7)

giving me back the error
"Error in plot.new():
! figure region too large
Backtrace:

1. base::plot(...)
2. graphics::plot.default(...)
3. graphics::plot.new()
   Execution halted"

Even if I reduce the plot height to 1 cm each which is obviouls too small. But still I get the error, the figure is too large.

3. only using par() to specify columns and rows:

par(mfrow=c(3,2))

plot(totDistM,OAM,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,OAW,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,PA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,PA_W,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,UA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,UA_W,pch = 21,col="black", bg="yellow", cex=0.7)

again getting very narrow plots like with attempt no. 1 (layout())

Up front, I recommend going with a native-faceting plot such as ggplot2 instead of staying with base graphics.

Starting vectors of data

I believe your data is merely a set of vectors in your local environment.

set.seed(42)
totDistM <- runif(20)
totDistW <- runif(20)
OAM      <- runif(20)
OAW      <- runif(20)
PA_M     <- runif(20)
PA_W     <- runif(20)
UA_M     <- runif(20)
UA_W     <- runif(20)
head(totDistM)
# [1] 0.9148060 0.9370754 0.2861395 0.8304476 0.6417455 0.5190959

I suggest this should likely be in a data.frame, either one big one or separated by M/W. More on that in a moment.

Base graphics

This can likely be remedied by compressing the top and bottom margins in the plot. The defaults of par(mar=c(5,4,4,2)+0.1) are wasteful when you have no title or x-axis labels, consider including par(mar = c(2,4,1,0) + 0.1). It'll help a little.

With just that change, we can see dramatic improvements:

nf <- layout(matrix(c(1,2,3,4,5,6), ncol=2,nrow=3, TRUE),widths = c(1,1), heights=c(1,1,1))
par(mar = c(2,4,1,0) + 0.1) # the only change to your code
plot(totDistM,OAM,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,OAW,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,PA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,PA_W,pch = 21,col="black",bg="yellow", cex=0.7)
plot(totDistM,UA_M,pch = 21,col="black",bg="orange", cex=0.7)
plot(totDistW,UA_W,pch = 21,col="black", bg="yellow", cex=0.7)

The order of the mar= numbers are in ?par:

     ‘mar’ A numerical vector of the form ‘c(bottom, left, top, right)’
          which gives the number of lines of margin to be specified on
          the four sides of the plot.  The default is ‘c(5, 4, 4, 2) +
          0.1’.

The 0 I have is for the right-side, clearly nothing to show there. The 2 is for the bottom, we need that to preserve the x-ticks; if you add an x-axis label then you'll need to increase this value (likely to 4 as on the left-side). You can reduce the 1 to 0 for the top to squeeze a little more space, assuming you have no title or subtitle.

Note: one direction people often go with plots like this is "I don't need x ticks/labels on every plot, only on the bottom two plots". This is correct, and not hard to do, just changing the arguments to individual calls to plot. However, note that layout(.) is going to preserve the overall side of each sub-plot regardless of the presence of axis ticks, so doing so will make some plots larger than others. (Similarly for y-ticks.)

nf <- layout(matrix(c(1,2,3,4,5,6), ncol=2,nrow=3, TRUE),widths = c(1,1), heights=c(1,1,1))
par(mar = c(0,4,2,0) + 0.1) # the only change to your code
plot(totDistM,OAM,pch = 21,col="black",bg="orange", cex=0.7, xaxt='n', main="totDistM")
par(mar = c(0,0,2,0) + 0.1) # the only change to your code
plot(totDistW,OAW,pch = 21,col="black",bg="yellow", cex=0.7, xaxt='n', yaxt='n', ylab='', main="totDistW")
par(mar = c(0,4,0,0) + 0.1) # the only change to your code
plot(totDistM,PA_M,pch = 21,col="black",bg="orange", cex=0.7, xaxt='n')
par(mar = c(0,0,0,0) + 0.1) # the only change to your code
plot(totDistW,PA_W,pch = 21,col="black",bg="yellow", cex=0.7, xaxt='n', yaxt='n', ylab='')
par(mar = c(2,4,0,0) + 0.1) # the only change to your code
plot(totDistM,UA_M,pch = 21,col="black",bg="orange", cex=0.7)
par(mar = c(2,0,0,0) + 0.1) # the only change to your code
plot(totDistW,UA_W,pch = 21,col="black", bg="yellow", cex=0.7, yaxt='n', ylab='')

Notice how the lower-left is the smallest, and the other vary. Can you work around this? Likely with a lot of playing, sure, but once the plotting canvas (all six combined) changes size, then other things happen and you will likely need to repeat to get the reshaping done just right.

(This is the main reason why I recommend ggplot2 or another plotting engine that handles these dimensions for you.)

On top of the use of par(mar=..), in the rmarkdown R chunk you can set the figure dimensions directly using fig.height or fig.dim in the chunk options, see https://bookdown.org/yihui/rmarkdown/tufte-figures.html.

ggplot2

This approach benefits from (and in some use-cases may require) the use of a data.frame in your data. Even if not required, it really makes many things much easier.

quuxM <- data.frame(totDistM, OAM, PA_M, UA_M)
quuxW <- data.frame(totDistW, OAW, PA_W, UA_W)
head(quuxM) # same for quuxW
#    totDistM        OAM      PA_M      UA_M
# 1 0.9148060 0.37955924 0.5816040 0.3567220
# 2 0.9370754 0.43577158 0.1579052 0.4106351
# 3 0.2861395 0.03743103 0.3590283 0.5734759
# 4 0.8304476 0.97353991 0.6456319 0.5896783
# 5 0.6417455 0.43175125 0.7758234 0.7196573
# 6 0.5190959 0.95757660 0.5636468 0.3949730

Now here it might get a little murky, but ... many plotting engines do much better with "long" data, especially when it comes time to faceting. I'll reshape/pivot it from the above "wide" format to a "long" format, handling the "X" columns separately from the "Y" columns. I'll also remove the M and W from the ends of the variables so that they can be compared side-by-side (trust me, see the pic at the end).

quuxMlong <- reshape2::melt(quuxM, id.vars = "totDistM", variable.name = "Ytype", value.name = "Y")
quuxWlong <- reshape2::melt(quuxW, id.vars = "totDistW", variable.name = "Ytype", value.name = "Y")
head(quuxWlong)
#     totDistW Ytype         Y
# 1 0.90403139   OAW 0.6756073
# 2 0.13871017   OAW 0.9828172
# 3 0.98889173   OAW 0.7595443
# 4 0.94666823   OAW 0.5664884
# 5 0.08243756   OAW 0.8496897
# 6 0.51421178   OAW 0.1894739

names(quuxMlong)[1] <- "X"
quuxMlong$Xtype <- "totDistM"
names(quuxWlong)[1] <- "X"
quuxWlong$Xtype <- "totDistW"
quuxlong <- rbind(quuxMlong, quuxWlong)
quuxlong$Ytype <- gsub("_?[WM]", "", quuxlong$Ytype)
head(quuxlong)
#           X Ytype          Y    Xtype
# 1 0.9148060    OA 0.37955924 totDistM
# 2 0.9370754    OA 0.43577158 totDistM
# 3 0.2861395    OA 0.03743103 totDistM
# 4 0.8304476    OA 0.97353991 totDistM
# 5 0.6417455    OA 0.43175125 totDistM
# 6 0.5190959    OA 0.95757660 totDistM

unique(quuxlong$Xtype)
# [1] "totDistM" "totDistW"
unique(quuxlong$Ytype)
# [1] "OA" "PA" "UA"

(There are other ways to get from your six vectors to a long frame. I'm not focusing on the options for this. Know that "wide form" frames make some things easier and "long form" frames make other things easier. In both cases, though, having frames is generally much better than work on lone vectors.)

From here, a single call handles all comparisons you had:

library(ggplot2)
ggplot(quuxlong, aes(X, Y)) +
  facet_grid(Ytype ~ Xtype, scales = "free") +
  geom_point(aes(fill = Xtype), shape = 21L) +
  scale_fill_manual(values = c(totDistM = "orange", totDistW = "yellow")) +
  labs(x = NULL, y = NULL) +
  theme(legend.position = "bottom")

With this, you have no need to try to collapse blank space between plots. Granted, ggplot2 does have a small learning curve from base graphics, but it brings with it many capabilities that require strict control and attention to detail in base graphics to be able to replicate safely and consistently.