ggplot中geom_pointrange处理OR值可视化及置信区间优化问题

Hey there! Let's tackle your ggplot OR plot issues step by step. First, I want to point out why converting your numeric OR/lower/upper values to factors caused the messed-up confidence intervals: geom_pointrange requires numeric values for y, ymin, and ymax—when you turn them into factors, ggplot treats them as discrete categories (each factor level gets a sequential integer position, like 1, 2, 3...) instead of their actual numeric values. That's why your confidence intervals were totally off.

Now let's get to your two desired goals:

Goal 1: Make the 0-1 interval clearer without distorting the scale (Recommended: Logarithmic Y-axis) #

OR values are perfect for logarithmic scales because they're ratio measures—this compresses the large upper bounds while expanding the space between 0 and 1, making all confidence intervals visible. Here's how to adjust your code:

library(ggplot2)

ggplot(dat, aes(x = region, y = or, ymin = lower, ymax = upper)) +
  geom_pointrange() +
  geom_hline(yintercept = 1, linetype = 2, color = "red") + # Highlight null OR value
  scale_y_log10(breaks = c(0.1, 0.5, 1, 2, 5, 10, 20, 50, 100, 300)) + # Custom readable breaks
  annotation_logticks(sides = "l") + # Add log-scale tick markers for clarity
  theme_bw() +
  theme(
    plot.margin = unit(c(1, 1, 1, 4), "lines"),
    axis.title = element_blank(),
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    legend.position = "none"
  ) +
  facet_wrap(~ country) +
  coord_flip()

Why this works: #

• The log scale stretches the 0-1 range so you can see narrow confidence intervals there (like region B in country A1, region B in country A2) without squishing larger values.
• annotation_logticks adds small secondary ticks to explain the log scale divisions, making interpretation easier for readers.
• Custom breaks ensure key reference values (0.1, 0.5, 1, etc.) are labeled, avoiding confusion.

Goal 2: Break the Y-axis to magnify the 0-1 interval #

If you prefer to keep a linear scale, you can use the ggbreak package to split the Y-axis into two segments: one magnified segment for 0-1, and another for the larger values. Here's how:

First install the package if you haven't:

install.packages("ggbreak")

Then modify your plot:

library(ggplot2)
library(ggbreak)

ggplot(dat, aes(x = region, y = or, ymin = lower, ymax = upper)) +
  geom_pointrange() +
  geom_hline(yintercept = 1, linetype = 2, color = "red") +
  scale_y_break(c(1, 5), scales = c(1, 0.5)) + # Break between 1 & 5; 0-1 segment is twice as tall
  theme_bw() +
  theme(
    plot.margin = unit(c(1, 1, 1, 4), "lines"),
    axis.title = element_blank(),
    axis.ticks.y = element_blank(),
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    legend.position = "none"
  ) +
  facet_wrap(~ country) +
  coord_flip()

Notes on this approach: #

• The scale_y_break function splits the axis at c(1,5), creating a 0-1 segment and a 5-350 segment. The scales argument controls relative sizing—c(1, 0.5) makes the 0-1 segment twice as large for better visibility.
• Axis breaks can sometimes be misleading to readers, so make sure to clearly note the break if you use this method in a report or presentation.

What not to do (and why) #

Your initial attempt to convert values to factors doesn't work because:

• Factors in ggplot are treated as discrete categories, not numeric values. When you set y = or (a factor), ggplot assigns each OR value a position based on its order in the factor levels (e.g., smallest value at y=1, next at y=2) instead of its actual numeric value.
• The ymin and ymax (also factors) get assigned the same integer positions, so confidence intervals don't align with the point estimates at all.

Hope these solutions help! Let me know if you need further tweaks.

内容的提问来源于stack exchange，提问作者acts