[LMB] OT: Gender Roles , figures

Matija Grabnar lmb at matija.com
Thu Sep 30 08:35:11 BST 2021

On 30/09/2021 03:32, Pouncer via Lois-Bujold wrote:
> Matija Grabnar suggests:
> >And your comparing the percentiles does nothing to address
> >the point made earlier that there is a significant overlap
> >in heights of the two populations.
> I object to the structure of the argument.
> Suppose I argue that there is no significant difference
> in "climate" between day and night.

Oh, I love the smell of a strawman in the morning.

Of course, climate is determined by long term temperatures and their 
averages, not single point measurements.

On the other hand, your argument was that height is a good proxy for 
strength, which is wrong. Or that weight is a good proxy for strength, 
which is even more wrong.

> You provide a
> record showing times of sunrise, sunset, time of high
> temperature and time of low temperature,  every day
> for the past year.  SON-of-a-GUN, day after day, the
> daytime high temps are higher than the overnight low
> temps.
> Then I come back and point out that over the full
> year, the "distributions have significant overlap"
> and there are some few nights in summer that are
> warmer than some days in winter.

Oh, you can't even can't even do your own strawman right. Over 
significant parts of the world, it is quite common for summer nights to 
be warmer than most winter days. The bell curves of daytime and night 
time temperatures in a given month are significantly more distinct than 
the bell curves of male and female strength.

> Anybody buy my analysis?  What makes the climate example
> different from the gender example?

1) You chose a poorly correlated proxy. That was the most significant error.

2) Even your strawman had to be biased with the "some few" wording, 
because it didn't reflect reality for significant portions of the globe.

3) And even if we accepted all your premises, there would be a 
temperature I could give you, and you couldn't guess if it was a night 
time or daytime temperature.

> Sylvia proposes:
> >If you say "Here's a 5"5' person, based on statistics
> >will they be male or female?"  you may as well flip a
> >coin as look at statistics.
> Well, Bayesian... Naah. Let's not.
> The chart I've already provided shows 89 percent of men
> are 5'5" or taller. Only 26 percent of women are that tall,
> or taller.

Ding! She didn't say "or taller". She said <this person is 5'5">, which 
is different, because you don't get to drag in "how much of the bell 
curve is to the right of this" into the calculation, you have to address 
the odds of sex based on a specific height. And that's with your lousy 

If the question was "this person can repeatedly move 40 pound things 
around all day long", determining the sex would be even harder. Which is 
the point we are trying to make you see.

More information about the Lois-Bujold mailing list