additional take on the israeli sperm count data

the really interesting issues may be in the outliers

this is a follow up piece to this one:

pfizer vaccine effects on total motile count in sperm donors

please read the 2 in conjunction.

in the swarm-sourced research world, you never get to do all the smart things. you miss stuff and then, once you see it, say “aha!” once someone shows it to you and it spurs you on to new thought.

and this is why it pays to have perceptive pals such as longtime gatopal™ and israeli data-maven ran israeli.

Ran Israeli @RanIsraeli

Seems like the authors took advantage of the Median vs Average differences to suggest that everything was fine 150 days after vaccination. Is it really so? Median disadvantages: "Median is not affected by very large/very small values." We need more data!

Ran Israeli @RanIsraeli

@EF20608797 Seems like the authors took advantage of the "Median vs Average" differences to suggest that everything was fine 150 days after vaccination. Median disadvantages - "The median is not affected by very large or very small values." In this specific case, small values matters! https://t.co/AECJoICVXJ

June 19th 2022

13 Retweets34 Likes

the implications of this finding are quite striking because median and average can be VERY different things.

median reverted to normal while average did not. mathematically, this implies/necessitates that the system has a few outliers driving the average.

we’re seeing a subset that got a bad outcome.

ran thoughtfully provided a useful graphic to explain how these measures differ.

i’d like to amplify it and apply it to the matter at hand:

let’s say we have 10 people each with a TMC of 10.

the average is 10. so is the median.

the two are interchangeable in a homogeneous population.

but now consider a population with an outlier:

9 have a TMC of 10, one drops to zero.

average now drops to 9 (10% drop).

but median is still 10.

drop a second person to zero, and average is now 8 (20% drop).

median is still 10.

(this would be about the same magnitude and outcome as the study shows)

so which do you want to trust for a medical outcomes study?

• median makes it look like nothing happened.

• but 1 in 5 people had their total motility count drop to zero. they were sterilized.

• that’s a helluva risk factor to ignore.

no measures, avg or median can tell you everything, but medians notoriously fail to capture subsets of outliers. that’s actually kind of why one uses them. but it also makes them inapt for studies of side effects in drugs as anything affecting fewer than half the cohort gets missed.

and that’s an awful lot to leave unexamined.

what we really need to see are the individual outcomes data. based on this avg/media divergence, i will wager it’s going to show us a severe drop in a few people that did not affect most.

we’ll see 20-40% of the group get deeply and durably suppressed while the rest experienced some lesser impact transitorily.

if somewhere on the order of 1 in 5 and 2 in 5 males are seeing severe, durable drops in TMC (50-100% drop), that’s a massive side effect profile. (4 in 10 dropping 50% has the same effect on the avg as 2 in 10 dropping 100%)

it also raises a number of questions about whether and to what extent this could be impairing other testicular function (like testosterone production) and this warrants study.

these are very important questions, especially if this is an autoimmune issue rather than just toxicity as that could well be irreversible and or cumulative with further dosing.

this is information we badly need in the public domain to make sound public health decisions.

in line with ran, i’d like to ask these researchers to release the full data so that we can make an assessment on that.