0
$\begingroup$

I have carried out some permutational sign tests, which give results in the following format:

                        Test1      Test2       Test3
Observed value           0.35       0.64        0.52
Expected range      0.41-0.51  0.49-0.61   0.47-0.55

I am happy with the results - tests 1 and 2 show observed values outside the expected range, while test 3 shows a non-significant difference.

I am struggling to think of an exciting way to display these results. Other people have suggested showing a histogram of the expected distributions, with arrows or lines showing that the observed results are outside (or inside) the range. However, I feel that the large histogram of expected values (which always shows a normal distribution) is essentially useless information - they are not "real" values, just results of randomisation.

I've been playing around with using a bar chart - with bars showing the observed values, and lines alongside showing the expected range. However, I'm having trouble making this look good in R - the lines always appear too separate or distant from the bars. And I'm not sure it will ever look that great.

Should I just give up and present the results as text? Or is there a better way to display such results? I don't see good visualisations of permutation tests very often.

Any ideas much appreciated!

$\endgroup$
1
$\begingroup$

Here's one go at it, although I'm not sure it counts as "exciting". I think better methods would make something out of the distribution over the expected range, not just the beginning and end points which give a false impression of fixedness. But this is probably better than a bar chart:

enter image description here

library(ggplot2)
library(ggrepel)

dat <- data.frame(test = 1:3, observed = c(0.35, 0.64, 0.52),
                  exp_min = c(0.41, 0.49, 0.47), exp_max = c(0.51, 0.61, 0.55))

# change the seed number below for slightly different configuration of "observed":
set.seed(126)
ggplot(dat, aes(y = test)) +
  geom_segment(aes(x = exp_min, xend = exp_max, yend = test), colour = "steelblue", size = 3, alpha = 0.5) +
  geom_point(aes(x = observed), size = 3) +
  geom_text_repel(aes(x = observed), label = "Observed") +
  scale_y_continuous("Test number", breaks = 1:3, minor_breaks = NULL) +
  labs(x = "Value\n(blue lines show expectation)")
$\endgroup$
  • $\begingroup$ Thanks, I found this useful inspiration. I also really appreciate you including the code - I haven't used ggrepe1 before, so it's useful to see the example. $\endgroup$ – rw2 Jan 29 '18 at 20:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.