Leukemia · Clin Cancer Res · 2012
The paper: chance would rarely make higher doses look this good. Its counts: chance easily could.
How often chance alone would produce a gap this big between dose groups
The paper says
On chemotherapy plus a new drug, 7 of 18 patients (39%) on its three higher doses went into remission, against 1 of 6 (17%) on the two lower. The paper reports p = 0.0005: chance alone would make that gap about 1 time in 2,000.
Its own data show
Its tables match. On those counts, the test the paper names gives p = 0.62: chance alone would make a gap this big about 6 times in 10. Even if all 8 remissions had been on the higher doses, p would be 0.066 (about 1 in 15). So 1 in 2,000 is out of reach.
A p-value is how often chance alone would produce a gap at least this big if dose made no difference. A group of 6 swings easily: one more remission there makes it 33% against 39%. A real dose gap would point to the new drug, since everyone got chemotherapy. These counts don’t show one. Its findings on safe doses stand.