Plotara guides
Practical guides for scientific analysis and figures
Clear, practical guidance for making defensible analyses and publication-ready figures.
How to add a number-at-risk table to a Kaplan-Meier plot
A number-at-risk table is expected in most clinical journals. What it shows, where it goes, and how to produce one without writing R.
EC50 and IC50 curve fitting without writing code
How to fit a dose-response curve, which model to choose, and the mistakes that produce a confident but wrong EC50.
Your two-way ANOVA has a significant interaction. Now what?
A significant interaction changes what your main effects mean. How to interpret it, what to report, and why simple effects come next.
Levene's test is significant. Which test should you use instead?
Unequal variances do not invalidate your analysis. Use Welch's ANOVA and Games-Howell instead of the standard F-test and Tukey.
How to report Welch's ANOVA and Games-Howell
The exact sentences, with decimal degrees of freedom explained, plus what to include for each pairwise comparison.
How to make a forest plot from a spreadsheet
What each element of a forest plot means, how to lay out your data, and how to produce one without writing R.
Journal figure width: 90 mm single column or 180 mm double?
Why journals specify figure widths in millimetres, how to choose, and how to size text so it survives the reduction.
How to add significance bars and asterisks to a plot
Where significance bars go, what the asterisk conventions mean, and why exact p-values are usually better.
Significant ANOVA, but no post-hoc comparison is significant
Why this happens, what it means, and how to report it without pretending you found something you did not.
Statistics software that exports Python code
Why exportable code matters for reproducibility, what to look for, and which tools give you a runnable script.
Mauchly's test is significant. Greenhouse-Geisser or Huynh-Feldt?
What sphericity means, which correction to apply, and how to report the adjusted degrees of freedom.
Shapiro-Wilk says your data are not normal. Do you have to switch tests?
A significant normality test does not automatically mean switching tests. What to check instead, and when it genuinely matters.
Correlation heatmap with non-significant cells masked
Why masking non-significant correlations makes a heatmap honest, and how to handle multiple comparisons across a correlation matrix.
How to build a multi-panel figure with A, B, C labels
Panel label conventions, how to keep panels visually consistent, and how to lay out a multi-panel figure at journal width.
Converting wide data to long format for plotting
What the difference is, why plotting and statistics tools want long format, and how to convert without writing pandas or tidyr code.
Repeated measures ANOVA or mixed ANOVA: which do you have?
How to tell which design you have, what each test assumes, and when to use a linear mixed model instead.
Pearson or Spearman: which correlation should you use?
The difference between linear and monotonic relationships, when each coefficient applies, and where Kendall fits in.
How to report Mann-Whitney U in APA format
The exact sentence structure, which statistics to include, and the effect size most people leave out.
Elbow plot: choosing k for k-means clustering
How to read an elbow plot, what to do when there is no obvious elbow, and which methods to use alongside it.
Exporting figures as SVG or PDF for journal submission
When to use vector and when to use raster, what DPI to export at, and the font problem that catches most people.
Your VIF is too high. What do you do about multicollinearity?
What VIF thresholds actually mean, when multicollinearity matters, and the fixes that work.
Your residual plot is funnel-shaped. What does that mean?
A fanning residual plot means heteroscedasticity. What it breaks, and the four fixes that work.
How to report a mixed ANOVA in APA format
The reporting order, what to do with a significant interaction, and how to handle the sphericity correction.
How to report a linear mixed model in a paper
What to include about the random effects structure, which statistics to report, and the details reviewers ask for.
Chi-square or Fisher's exact test?
The expected-count rule, why it exists, and which test to use for small samples and paired data.
Which effect size should you report for each test?
A direct mapping from test to effect size, what the benchmarks mean, and why confidence intervals matter more than the label.
What DPI do journals require for figures?
The standard DPI requirements for line art, halftones and combination figures, and why vector is better than any of them.
Colourblind-safe palettes for scientific figures
Which colour combinations to avoid, which palettes are safe, and why colour should never be your only signal.
Violin plot or box plot: which should you use?
What each one shows, when the extra detail of a violin is worth it, and why showing the points often beats both.
Paired data: connected points instead of bar charts
Why a bar chart of two means hides your paired design, and what to plot instead.