selection bias
#test-results 2025-02-06
Thursday34 messages6 participantstimes are UTC
Highlights from this day
i told WWB about a result they would not have liked and they did not argue. credit where it is due
one bad test does not condemn a supplier. one good test does not clear one. both halves get ignored equally
[edited]weigh it first
whats the lot
thats within range
if the vial has been through a bad transit you are testing the transit as much as the supplier
🧊3
right so post the lot, the purchase month, the lab, the number, and what you were expecting. that makes it useful later
lot bought purity assay/label
A-2418 2025-03 99.1% 9.6 / 10mg
A-2601 2025-08 98.7% 9.8 / 10mg
B-0114 2026-01 99.3% 9.4 / 10mg
B-0329 2026-04 98.9% 9.9 / 10mgmy vial assayed under label but the purity was fine, which matters more
quick one i test roughly every tenth vial. it is a sampling strategy, not paranoia
🔥5😂1
do you test every order or sample it
i have never regretted spending the money on a test. i have regretted not testing twice, for what its worth
one data point
how many lots do you need before you can say anything about a supplier
underfill is more common than impurity. if you only test purity you are testing the less likely failure
❤️1😂2👍5
coming back after 18 months with a follow up result, where do i put it
log the boring results too. a channel that only records failures produces a false picture
for the archive what interests me is when the gap goes the same direction every time. once is noise, three times is a pattern
*VendorInvestigate not the other one
attach the pdf
i had a result i am fairly sure was a handling error on my side and i logged it as such
update from 26 months ago: retested the same supplier, result came back better, logged both
thats a fill problem
😂1🤝1📈15
log the boring ones
Digest for the week of 2024-09-29 has been published.
🤝5🔥9
unrelated but quantity testing is usually a separate line item and it is the one worth paying for, i think
weigh the vial before you reconstitute. it costs nothing and catches the common failure, n of 1 obviously
the results in this channel skew because people test when they are suspicious. that is selection bias and it is real
97.4 against a certificate of 98.1 is fine. the same supplier landing low four times in a row is a question
a result without a lot number and a purchase date is an anecdote with a pdf attached
Batch lookup E-2205: 6 independent reports on file, earliest 2024-12-29.
👍12
a supplier disputing a result is rare here and it has never gone well for the supplier
if you are testing to settle an argument you will not settle it with n of 1
the widest spread in my data is about 2.6 percentage points across five lots from one supplier