alluvial box label is cut-off

[epo071]

this graph shows best the issue with the lable' Academy' being cut on half height.

[mbojan]

OK, I see.

Is Academy essentially an empty category or you are adding it separetely with mtext or something like that?

I think I'll expose mar as an additional argument.

This is a duplicate of #31

[mbojan]

I have added mar argument to alluvial() in https://github.com/mbojan/alluvial/commit/dffa0cc7d9b787c9aaeb77fdcbd83f02704c4833. Please try installing from GitHub with devtools::install_github and check if you can modify the margins as you like.

[epo071]

'Academy' is neither an empty category, nor mtext. The issue must rather do with positioning the block label within the plot limits. btw your latest commit also seems to break the colors.

[colinbrislawn]

Hello @epo071,

There is a package called ggalluvial based on this package, which has some more flexibility in labels styles. Check out the examples here: https://github.com/corybrunson/ggalluvial

[mbojan]

@colinbrislawn it is not an issue of inflexible labeling.

@epo071 I'm trying to reproduce the problem. It does seem to me that you have basically a row with frequency 0 (that's what I meant as "empty categories"), e.g.:

d <- data.frame(
  x = 1:3,
  y = 1:3,
  n = c(10, 20, 0)
)
alluvial(d[,1:2], freq=d$n, col="lightskyblue", border="white", blocks=FALSE)

Filter-out the rows with 0 frequencies and it should be OK.

PS. Thanks for the heads-up with the colors. I have reverted a recent PR that was bringing otherwise attractive enhancement of being able to color individual segments. It broke the col specification unfortunately.

[epo071]

here comes the reproducible data proof that the Academy or Tech. Coope boxes are NOT zero

PC_long <- data.frame(PRODUCTS= c('Search','Examination','Opposition','Tech. Coope','Academy','Filing', 'Appeal', 'Pat. Info','Search', 'Examination','Filing', 'Opposition','Appeal', 'Pat. Info','Tech. Coope','Academy','Search', 'Examination','Pat. Info','Tech. Coope','Filing', 'Appeal', 'Opposition','Academy','Examination','Filing', 'Search', 'Opposition','Pat. Info','Appeal', 'Pat. Info','Academy','Search', 'Examination','Tech.Coope','Filing','Opposition','Tech. Coope','Search', 'Examination','Academy','Filing', 'Opposition','Appeal', 'Pat. Info','Academy'),
MCC= c('JC', 'JC', 'JC', 'JC', 'JC', 'JC', 'JC', 'JC', 'Other', 'Other', 'Other', 'Other',      
'Other', 'Other', 'Other', 'Other', 'IM', 'IM', 'IM', 'IM', 'IM', 'IM', 'IM', 'IM',  
'SIS', 'SIS', 'SIS', 'SIS', 'SIS', 'Appeal','Pat. Info','Pat. Info','Pat. Info','Pat. Info','Pat. Info','Pat. Info',
'Pat. Info','Tech. Coope','Tech. Coope','Tech. Coope','Tech. Coope','Tech. Coope','Tech. Coope','Tech. Coope','Tech. Coope','Academy'),
value= c(592.553,223.780,31.372,1.490, 1.079, 1.024, 0.332, 0.287,  143.370, 75.176, 26.264,   15.849,   12.835, 9.988, 5.989, 1.881,  104.073, 22.627 ,  20.968, 15.795, 12.730, 3.237, 3.005, 0.432, 30.821, 18.190, 12.092, 9.128, 7.292, 56.435, 20.229, 0.799, 0.399, 0.155, 0.108, 0.022, 0.021, 13.285, 0.559, 0.311, 0.152, 0.107, 0.105, 0.088, 0.034, 6.396))

PC_long <- PC_long %>% 
  dplyr::arrange(MCC, desc(value)) %>% 
  dplyr::mutate(color= ifelse(MCC == 'JC', "pink", "lightskyblue")) %>% 
  dplyr::mutate(color= ifelse(MCC == 'IM', "purple", color))

alluvial::alluvial(PC_long[,rev(1:2)], freq= PC_long$value,
                     col = PC_long$color,
                     blocks= FALSE,
                     ordering= list(NULL,order(PC_long$MCC)),
                     hide=PC_long$value < 13)

[mbojan]

Thanks for the example. The clipping is due to ylim passed to plot. If the top or bottom stripes have low frequency the labels extend beyond the plotting region. I think I'll expose xlim and ylim as arguments, or at least allow for an offset that will modify the default values.

[mbojan]

I added xlim_offset and ylim_offset to adjust the limits of the plotting region. Now e.g. with

alluvial::alluvial(PC_long[,rev(1:2)], freq= PC_long$value,
                   col = PC_long$color,
                   blocks= FALSE,
                   ylim_offset= c(-0.05, 0.05),
                   ordering= list(NULL,order(PC_long$MCC)),
                   hide=PC_long$value < 13)

creates more space vertically so there is no border clipping.

In general in situations in which you have low-frequency stripes the clipping (also for labels of low-frequency categories next to each other) can be avoided by

• creating a plot with a different aspect ratio, i.e. larger vertically than horizontally
• increasing the inter-category gap.width

[epo071]

Thanks a lot indeed, issue is resolved from my point of view.