Paperonium

Lupus and memory · BMC Med · 2013

The summary prints a failing 0.065 yet says “significantly”; the paper’s own results: a passing 0.015.

1 in 100every time1 in 20the usual barabout1 in 15The summary: failsabout1 in 67Its main text: passes
1 in 100every time1 in 20the usual barabout1 in 15The summary: failsabout1 in 67Its main text: passes

How often chance alone would give this gap if the antibody did nothing

The paper says

Mice were injected with a normal antibody or a lupus-linked one. A mouse that remembers a maze spends extra time in a new branch. The opening summary: normal-antibody mice spent “significantly” more extra time there, 42% versus 9%, with a chance score (p) of 0.065.

Its own data show

The main text says 46% versus 9%, p = 0.015; the chart’s caption says 0.02, 0.015 rounded. Reading that chart’s bars (groups of 10 and 11 mice) and redoing the paper’s calculation gives 0.015 again. Even the summary’s 42% gives 0.025 to 0.034, still passing. With the chart’s numbers, 0.065 needs groups of 2–3 mice.

The chance score is roughly how often chance alone would give a gap this big if the antibody did nothing. The paper counts under 0.05, 1 time in 20, as “significant”: 0.065 (1 in 15) fails, 0.015 (1 in 67) passes. So the result most likely passes, but the summary, the most-read part, prints a failing number.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

16/6-idiotype expressing antibodies induce brain inflammation and cognitive impairment in mice: the mosaic of central nervous system involvement in lupus · BMC Med (2013) · doi:10.1186/1741-7015-11-90