undefined | Better HN

0 pointsslg5y ago0 comments

>I can think of contexts, especially when n=15(if this was a statistical study on insulin response to artificial sweeteners, people would be saying the sample size is too small to draw conclusions), but I don't know if developing hypothetical scenarios is useful.

N = 15 while p ≈ 20. It has been a while since I have taken a stats class, but that sample seems plenty large enough to me.

>Learning the true context is most important. This could be done by corroborating allegations for example.

Corroborating is exactly what journalists do. From the article:

>five people with knowledge of the situation said.

>But according to 23 current and former Coinbase employees, five of whom spoke on the record, as well as internal documents and recordings of conversations

>according to a recording of the session shared with The New York Times

>In a company email he sent later, which was also shared with The Times

>wrote in a Slack message that was viewed by The Times.

>three people briefed on the situation said

>according to a recording of the event

>according to a copy of the message seen by The Times

>according to a copy reviewed by The Times.

>two people with knowledge of the situation said

The NYT talked to dozens of people, watched/listened to multiple recordings, and viewed numerous emails and Slack messages. This story is corroborated.

0 comments

opinion-is-bad5y ago

Addressing only the viability of the sample for basic binomial analysis, the sample size is on the small end. This would be okay if we could reasonably expect the underlying events to follow identical and independent distributions, but when we know the people involved worked together that assumption is clearly void. Unfortunately this means we really can’t use any sort of statistical analysis on this datum. The models just don’t support it.

oh_sigh5y ago

If you order the claims in the article on a scale of not racist at all to clearly racist, you'll see basically all of the corroboration comes from claims on the left side of that scale, and few if any on the right side. The claims I was specifically thinking about were things like:

> One Black employee said her manager suggested in front of colleagues that she was dealing drugs and carrying a gun, trading on racist stereotypes. Another said a co-worker at a recruiting meeting broadly described Black employees as less capable. Still another said managers spoke down to her and her Black colleagues, adding that they were passed over for promotions in favor of less experienced white employees. The accumulation of incidents, they said, led to the wave of departures.

These have no corroboration, even though they allegedly happened in front of multiple coworkers.

slgOP5y ago

We don't need to corroborate each individual complaint. The story here isn't that one employee was passed over for a promotion. The story is the trend that all these anecdotes support. There is corroboration for the trend.

googlryas5y ago

But the point is that the corroborations are all on the lesser or non-complaints of the article, not the most egregious ones. To make an extreme example, posting a series of actual facts that culminate in aliens have visited earth doesn't make the latter assertion any more true.

edit: Sorry, I'm the poster from above, I just posted from my phone which has a different account logged in that I originally intended to give hiring advice unlinked to my main account(opsec fail).

1 more reply

inglor_cz5y ago

Interestingly, "carrying a gun" is an American stereotype here in Europe, not race-specific.

Isn't there like 400 million guns in private hands in the US? If so, the assumption that pretty much anyone you meet in the street might have concealed gun on them is not absurd per se.

But speaking about it aloud seems strange, yes.

"passed over for promotions"

This is the oldest workplace complaint ever and happens in racially homogeneous countries all the time. Favoritism in the workplace is probably as old as the pyramids.

j / k navigate · click thread line to collapse