orchard_plot() not robust to moderators with non-capitalised categories

[befriendabacterium]

Describe the bug
When the levels of a categorical moderator are not capitalised to begin with, orchard_plot() breaks with changes to 'upper' or 'tree.order' options. Below I demonstrate with a reproducible example where I turn the factor to lowercase first. I also provide a workaround for the 'tree.order' break that might help you come up with a fix.

To Reproduce

# NON-CAPITALISED CATEGORIES --------------------------------------------------

data(lim)

#assume phylum is a lower-cased categorical variable from the start
lim$Phylum<-tolower(lim$Phylum)

# Add in the sampling variance
lim$vi <- (1/sqrt(lim$N - 3))^2

# Lets fit a meta-regression - I will do Article non-independence. The
# phylogenetic model found phylogenetic effects, however, instead we could fit
# Phylum as a fixed effect and explore them with an Orchard Plot
lim_MR <- metafor::rma.mv(yi = yi, V = vi, mods = ~ Phylum, random = list(~1 |
                                                                            Article, ~1 | Datapoint), data = lim)
summary(lim_MR)

## UPPER = F (WORKS) ------------------------------------------------

# Plot the meta-regression model with default order
orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE, 
             upper = T)

## UPPER = F (BREAKS) ------------------------------------------------

orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE, 
             upper = F)

## CHANGE ORDER (BREAKS) ------------------------------------------------

new.order<-rev(sort(unique(lim_MR$data$Phylum)))

# Plot the meta-regression model with new order
orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE,
             tree.order = new.order)

## WORKAROUND FOR CHANGING ORDER BUG ------------------------------------------------

#Apply firstup to new.order to get it to work (workaround)
new.order_firstup<-firstup(new.order)

# Plot the meta-regression model with new order
orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE,
             tree.order = new.order_firstup)

Expected behavior

# CAPITALISED CATEGORIES --------------------------------------------------

data(lim)

# Add in the sampling variance
lim$vi <- (1/sqrt(lim$N - 3))^2

# Lets fit a meta-regression - I will do Article non-independence. The
# phylogenetic model found phylogenetic effects, however, instead we could fit
# Phylum as a fixed effect and explore them with an Orchard Plot
lim_MR <- metafor::rma.mv(yi = yi, V = vi, mods = ~ Phylum, random = list(~1 |
                                                                               Article, ~1 | Datapoint), data = lim)
summary(lim_MR)

## UPPER = T (WORKS) ------------------------------------------------

# Plot the meta-regression model with default order
orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE, 
             upper = T)

## UPPER = F (WORKS) ------------------------------------------------

orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
             alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE, 
             upper = F)

## CHANGE ORDER (WORKS) ------------------------------------------------

new.order<-rev(sort(unique(lim_MR$data$Phylum)))

# Plot the meta-regression model with new order
orchard_plot(lim_MR, mod = "Phylum", group = "Article", xlab = "Correlation coefficient",
                      alpha = 0.5, transfm = "tanh", angle = 45, N = "N", cb = FALSE,
                      tree.order = new.order)

Desktop (please complete the following information):

• OS: Windows 10
• Browser: NA
• Version: NA

Additional context
Add any other context about the problem here.

[daniel1noble]

@befriendabacterium Thanks for this mate! So sorry I didn't see this. I can't understand why I'm not getting notified when there's a new issue. I have all the notifications setup to be received (or so I thought). I'll see if I can find some time to fix this, but feel free to chuck in a pull request as this looks fairly simple

[befriendabacterium]

Hi Dan - no worries, not sure why that is. I rely on the email notifications to be honest so only thing I can suggest is make sure they're not going to your junk?!

I don't have a fix for this right now, just the workaround above sorry! Bit hectic at the mo to give this the brain power it needs, but I hope it's an easy fix.

Thanks!
Matt