Understanding Error-Bar Plots

When to use an error-bar plot

Error-bar plots are useful when you want a compact summary of group-level values for many cytokines at once. In CytokineProfile Shiny, they are most helpful when your question is:

how do group summaries compare across many cytokines?
do the group means or medians move in a consistent direction?
which cytokines look most different before deeper follow-up?

This method is best when a clean summary view is more useful than showing every raw point.

When not to use an error-bar plot

Error-bar plots are usually not the best first choice when:

the raw distributions are important and should be shown directly
the dataset is small enough that hiding individual points would lose too much context
you need a formal differential-summary method such as a volcano plot

If you want to see full distribution shape, Understanding Boxplots and Violin Plots is usually a better companion.

Example context

A typical use case is comparing treatment groups or responder groups across many cytokines after Step 2 filtering has defined the cohort and comparison variable.

What the app is showing

The error-bar workflow summarizes each cytokine within each group using a selected center and uncertainty metric, then optionally adds significance annotations across groups.

The figure answers a summary question first: how do the group-level values compare across the measured variables?

Which Step 4 arguments matter most

The highest-value controls are:

Comparison Column: the grouping variable being compared.
Statistic: whether the bar height represents the mean or median.
Error Metric: what the error bars represent.
Statistical Test: how the significance annotations are computed.
Multiple-Testing Correction: whether p-values are adjusted across many comparisons.
Number of Columns: how many cytokines are shown per page.

Interpret these settings carefully:

Statistic = Mean pairs naturally with Standard error, Standard deviation, or 95% CI.
Statistic = Median is often easier to trust when the data are skewed or contain outliers.
95% CI and Standard error describe uncertainty more than raw spread.
Standard deviation and MAD describe variability in the measurements themselves.

How to read the main output

A useful reading order is:

Identify the cytokines where group summaries are farthest apart.
Check whether the chosen Error Metric suggests stable or noisy group estimates.
Read any statistical annotations only after remembering which Statistical Test and correction were used.

Interpretation checklist:

large bar differences are more persuasive when the uncertainty bars are not overwhelmingly wide
overlapping error bars do not automatically mean “not significant”
non-overlapping error bars do not automatically mean “significant”
significance markers should be read as summaries of the selected test, not as replacements for distribution inspection

Common cautions

Important cautions include:

error bars are often misread, especially when users forget whether they show spread or uncertainty
raw data shape can be hidden, including skewness, bimodality, and outliers
significance annotations can look stronger than the underlying effect size really is
many side-by-side cytokines can make the page visually dense, so Number of Columns matters

How to reproduce the result in the app

Filter the dataset to the groups and cytokines you want to summarize.
Choose Error-Bar Plot.
Set Comparison Column.
Choose Statistic and Error Metric.
Review Statistical Test and Multiple-Testing Correction if you want annotations.
Adjust Number of Columns if the figure feels crowded.
Use boxplots or violin plots as a follow-up when the summary bars are not enough.