@tomhfh Tom. Promise me you will never teach statistics.
The graph you have posted clearly shows two overlapping normal distributions.
Each normal distribution is associated with either the female sex or the male sex.
The graph you have posted clearly shows two overlapping normal distributions.
Each normal distribution is associated with either the female sex or the male sex.
@tomhfh As you correctly point out, short males are not female.
Yet a very short male may appear in the little area of overlap highlighted, because they are at the far left of the male normal distribution, not because they are magically ‘intersex’ or ‘a bit female’.
Yet a very short male may appear in the little area of overlap highlighted, because they are at the far left of the male normal distribution, not because they are magically ‘intersex’ or ‘a bit female’.
@tomhfh The X axis in the graph is not ‘sex units’. The graph is not mapping sex. It is a mapping schematically a characteristic associated with sex, like testosterone levels (in some concentration unit).
Sex is why you have a bimodal distribution of testosterone levels.
Sex is why you have a bimodal distribution of testosterone levels.
@tomhfh Furthermore, please tell me how you create quantitative distributions of categorical data like ‘karyotype’ (chromosomes). There is no intrinsic order to a dataset like XX, XY, XO, XYY etc.
In what units do you measure karyotype? 🤦♀️
In what units do you measure karyotype? 🤦♀️
@tomhfh To repeat, Tom. The reason you get bimodal sex characteristics is because you are measuring variable characteristics from two discrete populations that have an average difference in the said characteristic.
@tomhfh Apologies. The male distribution is in the left, so a very short male might appear on the far right of this distribution.
@tomhfh Let me illustrate.
Measure the body length of 1000 domestic rabbits and 1000 domestic dogs.
Plot them as length distributions. A tiny dog may well be within the ‘rabbit length’ range.
A tiny dog is not, in fact, a rabbit. You are not mapping a dog-rabbit species spectrum.
Measure the body length of 1000 domestic rabbits and 1000 domestic dogs.
Plot them as length distributions. A tiny dog may well be within the ‘rabbit length’ range.
A tiny dog is not, in fact, a rabbit. You are not mapping a dog-rabbit species spectrum.
@tomhfh You are mapping a variable characteristic (body length) in two discrete populations (dogs and rabbits), where said characteristic may, at the more extreme converging ends of each individual distribution, overlap in value.
@tomhfh If you choose to present both of those distributions on the same graph, perhaps to compare them, you don’t suddenly eradicate the *very first premise* that you have collated the data from two discrete populations.
@tomhfh In fact, it’s deeper than that.
Say you don’t know you are collecting data from two discrete populations - that is, you were given a dataset called ‘body length of domestic pets’ with 2000 entries.
Say you don’t know you are collecting data from two discrete populations - that is, you were given a dataset called ‘body length of domestic pets’ with 2000 entries.
@tomhfh If you plots all those body lengths and identify a bimodal distribution, the very first question you should be asking is: am I looking at data from two different types of pet?
Stats 101, Tom.
If you see a trimodal distribution, maybe you’ve got goldfish data in there.
Stats 101, Tom.
If you see a trimodal distribution, maybe you’ve got goldfish data in there.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
