# Constructed data: grip strength (g) in 20 mice, 5 per sex per treatment.
# True model: males 20 g stronger at baseline; the drug adds 12 g in BOTH sexes; SD 12.
# The seed (362) was chosen from the first 400 so that, as often happens, the per-sex t-tests disagree
# (significant in males, not in females) although the true effect is identical (see search_seed.py).
import numpy as np, pandas as pd
rng = np.random.default_rng(362); rows = []
for sex, base in [("female", 100), ("male", 120)]:
    for trt, eff in [("vehicle", 0), ("drug", 12)]:
        for i in range(5):
            rows.append((f"{sex[0].upper()}{trt[0]}{i+1}", sex, trt, round(base + eff + rng.normal(0, 12), 1)))
pd.DataFrame(rows, columns=["mouse", "sex", "treatment", "grip_g"]).to_csv("grip.csv", index=False)