By Yishan Mai

A difficult question I ran into early in my PhD was: When multiple experimenters do the same experiment but produce different results, what do you do?

Many would either cherry-pick the “best” replicate or blindly average results; the first conceals data while the second is statistically unsound. To solve this problem, we developed mini meta-analysis for DABEST 2.0, which lets you synthesize results from internally replicated experiments. It allows you to:

— Visualize effect sizes from each replicate
— Compute a weighted meta-analytic effect
— See the consistency (or heterogeneity) across your replicates

So next time you and your colleagues have the urge to argue on whose replicate is more “correct”, consider using mini meta-analysis to combine your data into a single, meaningful conclusion, while maintaining transparency in data reporting.

Preprint: https://doi.org/10.64898/2026.01.26.701654
Code: https://github.com/ACCLAB/DABEST-python

Work in collaboration with: Zinan Lu, Jonathan Anns, Sangyu Xu, Nicole Lee, Hyungwon Choi, Adam Claridge-Chang, and others.

[Also posted here.]

By Jonathan Anns
​
Why your repeated-measures data deserves better than a simple line

Whenever the same subjects are measured more than once, whether across timepoints, doses, or conditions, you have a repeated-measures design. It's one of the most common frameworks in biomedical research. Yet research papers typically reduce these experiments into a mean line with error bars and P-values (Fig. 1A).

So, what's missing? The individual trajectories. The sense of variability. The actual magnitude of change.

In addition, the typical analysis approach entails a combinatorial explosion of post-hoc tests computing every possible pairwise comparison (Fig. 1B), many of which are not relevant to your hypothesis and unnecessarily inflate your multiple comparisons burden. The questions that actually motivated the study (when does the effect begin?, how large does it grow?, and does it persist?) are obfuscated.

Our new software, DABEST 2.0, is designed around a different approach: keep the individuals, and draw the overall shape. Our repeated-measures figure has two panels doing two distinct jobs. The upper panel shows observed values and their dispersion, an attribute of the sample (Fig. 1C). The lower panel shows the bootstrap distribution of the effect at each timepoint, an inference of precision that sharpens as the sample grows (Fig. 1D).

With DABEST 2.0, you can:
— Visualise each subject's individual trajectory alongside the means.
— Report the effect size with confidence intervals for the comparisons you care about.
— Show the full distribution of differences.

The result is a figure that more clearly quantifies (in a pretty way!) what changed, for whom, and by how much.

Preprint: https://doi.org/10.64898/2026.01.26.701654
Code: https://github.com/ACCLAB/DABEST-python

Work in collaboration with: Zinan Lu, Yishan Mai, Sangyu Xu, Nicole Lee, Hyungwon Choi, Adam Claridge-Chang, and others.
​

Here's a dirty secret about ANOVA: it tests a null hypothesis that nobody cares about. When you run a one-way ANOVA, you're testing whether "all group means are equal." But even if you reject this hypothesis, you learn nothing about which groups differ, in which direction, or by how much.

So you embark on a second analytical step: multiple two-group comparisons. A modest six-group experiment suddenly requires testing 15 hypotheses. To manage this multiplicity, you apply corrections like Bonferroni, which undermine your statistical power. What you posed as a focused research question has sprawled into a complex web of subsidiary tests, forced by the ANOVA ritual.

Our new preprint, "Getting over ANOVA: Estimation graphics for multi-group comparisons," makes the case for a better approach. Estimation statistics encourages you to compare each test group to a single control, focusing on the effect sizes that actually matter. A six-group experiment focuses attention on just five effect sizes with confidence intervals, showing magnitude and precision directly.

The preprint introduces estimation methods for a range of multi-group designs: repeated-measures experiments, 2×2 factorial designs, binary outcome data, and mini-meta analysis for internal replicates. Each can replace data-analysis practices used in thousands of studies every year.

Read our new preprint here: https://doi.org/10.64898/2026.01.26.701654

Also posted on LinkedIn. #Statistics#OpenScience#DataVisualization #Research

I was writing about serotonin-receiving neurons and reached for "serotonoceptive." The word should exist, but it doesn't.

We have "dopaminergic" for neurons that release dopamine, so why no equivalent for neurons that receive it? Instead, the literature is full of workarounds: "dopamine-sensitive neurons," "neurons expressing dopamine receptors," "dopamine target cells."

A solution was hiding in plain sight: "nociceptive" and "proprioceptive" have been around since Sherrington. Recent papers already use "GABAceptive" and "dopaminoceptive."

So I wrote a short paper proposing we generalize the '-ceptive' suffix. Dopaminergic neurons release dopamine; dopaminoceptive neurons receive it. Simple, systematic, and searchable.

Read the editorial here: https://doi.org/10.5281/zenodo.18373728