Thread by @FondOfBeetles

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie I'm not entirely sure where to start here.

"The average adult man differs from the average adult woman in a number of respects. Average height is taller..."

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie That's true. But height in humans is not bimodal. If you measure the entirety of the population, you get a normal distribution. It's only once you disaggregate by sex that you reveal any bimodality (although even that separation doesn't meet the criteria for true bimodality).

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie From Schilling et al, 2002, mythbusting the premise that human height is bimodal.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie "Average strength is greater".

Again, true. But again, I doubt sampling the population as a whole would reveal a bimodal distribution. Only disaggregation by sex allows us to understand that we have two overlapping populations.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie From the NHANES database, compiled by a Reddit user (grasshoppermouse).

If there were no male/female separation of colour and no mean lines, you really think grip strength would be bimodal across the population?

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie So how can we determine that males are taller and stronger on average? What reference characteristic do we anchor to?

Plotting secondary characteristics as if they were able to predictively discriminate the sex of someone is arse backwards.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie We know that males are taller that females because we have measured height in males and in females, and we have identified males and females using a feature that isn't height.

What features might we use?

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie Testosterone is more likely a genuine bimodal. You'll have around half your sample clustering in a narrow peak close to zero, then the other half of your sample somewhere up the graph clustering more widely around a much more shallow second peak.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie From Stanton, 2011.

The graph is coloured by sex, but even if it weren't, it's obvious that it's a bimodal distribution.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie Bimodality is almost always underpinned by two discrete categories/groups.

What might the discrete categories/groups be?

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie As you start measuring parameters that are a conceptual distance from the discrete category division, the data spread around each peak widens, as the parameter becomes less strongly influenced by the categorical division.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie This is why height is not bimodal at the population level. Although we know it is affected by sex, and we can pull out clear differences when we disaggregate by sex, it is also influenced by other factors like nutrition, ethnicity, etc.

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie Testosterone is a stark bimodal - this tells us that testosterone level is more directly influenced by the discrete category division.

What discrete category division underpins testosterone level?

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie You've spent a lot of posts describing what parameters might contribute to your swarm distributions (and got a few wrong), and just one sentence where you skip over the discrete category parameter that underpins all of your projected measurements.

"People with small gametes".

@threadreaderapp @ElizabethLidd @Phildidgee @Cave_Art_Films @ZombieTron @Sceptical_Woman @JanKaas0 @KatyMontgomerie You're too busy plotting starbursts of dots and mapping everyone to some unique composite "score" to even ask "why are there are two clusters?"

Work down to basic principles, you'll see how from that basic principle all of your starbursts explode.