← Back to Simulations

T-Cell Count: Two-Group Comparison

A control group and a treatment group have different median T-cell counts — but with only a few dozen subjects each, could a gap this big show up even if the treatment did nothing?

Data: T-cell count per unit area for two groups — Control (A, n = 49) and Treatment (B, n = 45).

Looking at the two groups

Each dot is one subject. Treatment (B) sits visibly higher than Control (A) — but "visibly higher" isn't yet a number. Toggle the box below to see each group's mean, MAD (median absolute deviation), and median.

Beeswarm comparison

Toggle box + stats

The same picture, shown as overlapping distributions instead of individual points: histograms first, then smooth kernel density curves on top.

Histogram & KDE comparison

Toggle KDE curves

Testing the gap: the big-box method

The idea. If treatment didn't matter, then Control and Treatment are really just two random samples from the same underlying population — so it shouldn't matter which subjects we call "A" and which we call "B". Pool everyone into one big box, then draw two brand-new samples from it — the same sizes as before (49 and 45), with replacement. Compute the gap between their medians. Do that thousands of times, and you get the distribution of gaps you'd see from pure chance alone.

Big-box resampling test

Pooled, then resampled with replacement

First watch Control and Treatment physically merge into one pooled box — then draw two fresh samples from it. The gap between their medians joins the null distribution below.

Reading the result: the observed gap (Δobs ≈ 9.7) sits far out past nearly every simulated draw from the pooled data — a gap this large essentially never happens by chance alone, which is why the two-group difference here is considered highly significant.