10.2 Statistics
213
The coefficient is usually denoted by r . Suppose that a result is r = 0.9; how
should you interpret that? If you take the square then you get r = 0.81 and
that seems further away from unity. But if you take the square root then you
get r = 0.95; that seems to be a much better correlation. So, Kapteyn wanted
to know, what is the ‘most natural measure of correlation’? Quite formal and
subjective. It hardly seems like an interesting question, but Kapteyn had the
opinion, which we now do not share with him anymore, that there should be
something of a causal relationship in the correlation. In the end Kapteyn did
come to the conclusion that it is best to simply take r . The definition of r in
mathematical terms has been given in Appendix A.7.
But there was a catch. The quantity r was introduced by Pearson! Although
it was based on Galton’s work, the quantity is officially called the ‘Pearson
product moment correlation coefficient’. Now it is a fact that Kapteyn did not
even mention the name Pearson! He did refer to Auguste Bravais, whom we
met earlier in his study of the Apex motion of the Sun, and treated him as
the founder of the study of correlation. It is not known if Pearson has seen
this article by Kapteyn; as mentioned, it was published in a British, albeit
astronomical, journal.
Kapteyn wouldn’t give up easily in this kind of disagreement about priority
and honor. In 1906, he approached David Gill and ask him to mediate to
get an article on this matter in the renowned journal Philosophical Transactions
of the Royal Society. Kapteyn believed that ‘biologists under the influence of
Karl Pearson were on the wrong track’ with their study of skewed distributions.
Even better had been the journal Biometrika, but Pearson was its general editor.
David Gill wrote that he was prepared to do what Kapteyn asked, but in the
end Kapteyn apparently did never manage to actually write such an article.
Kapteyn wrote to David Gill in 1907 that due to lack of time he would not be
coming to Leicester, where David Gill, as President of the British Association,
would give a lecture in which he wanted to speak prominently about Kapteyn,
and for which he had been informed extensively about his work (see Sect. 6.5).
Kapteyn mentioned in a letter that his involvement in statistical matters had
got quite out of hand and wondered what kind of evil spirit had prompted him
to do so. David Gill reacted by emphasizing that a gifted man like Kapteyn
should not be concerned with things that others could look after equally well.
Finally, what is the value or importance of a frequency curve in relation to the
structure of the Universe? I am happy that you admit there is an evil spirit
involved. Some form of exorcism is necessary and I would like to apply it.
In spite of this, Kapteyn kept himself busy with statistics, probably partly
in a sincere attempt to help his biological friends, but his stubborn desire to
213
The coefficient is usually denoted by r . Suppose that a result is r = 0.9; how
should you interpret that? If you take the square then you get r = 0.81 and
that seems further away from unity. But if you take the square root then you
get r = 0.95; that seems to be a much better correlation. So, Kapteyn wanted
to know, what is the ‘most natural measure of correlation’? Quite formal and
subjective. It hardly seems like an interesting question, but Kapteyn had the
opinion, which we now do not share with him anymore, that there should be
something of a causal relationship in the correlation. In the end Kapteyn did
come to the conclusion that it is best to simply take r . The definition of r in
mathematical terms has been given in Appendix A.7.
But there was a catch. The quantity r was introduced by Pearson! Although
it was based on Galton’s work, the quantity is officially called the ‘Pearson
product moment correlation coefficient’. Now it is a fact that Kapteyn did not
even mention the name Pearson! He did refer to Auguste Bravais, whom we
met earlier in his study of the Apex motion of the Sun, and treated him as
the founder of the study of correlation. It is not known if Pearson has seen
this article by Kapteyn; as mentioned, it was published in a British, albeit
astronomical, journal.
Kapteyn wouldn’t give up easily in this kind of disagreement about priority
and honor. In 1906, he approached David Gill and ask him to mediate to
get an article on this matter in the renowned journal Philosophical Transactions
of the Royal Society. Kapteyn believed that ‘biologists under the influence of
Karl Pearson were on the wrong track’ with their study of skewed distributions.
Even better had been the journal Biometrika, but Pearson was its general editor.
David Gill wrote that he was prepared to do what Kapteyn asked, but in the
end Kapteyn apparently did never manage to actually write such an article.
Kapteyn wrote to David Gill in 1907 that due to lack of time he would not be
coming to Leicester, where David Gill, as President of the British Association,
would give a lecture in which he wanted to speak prominently about Kapteyn,
and for which he had been informed extensively about his work (see Sect. 6.5).
Kapteyn mentioned in a letter that his involvement in statistical matters had
got quite out of hand and wondered what kind of evil spirit had prompted him
to do so. David Gill reacted by emphasizing that a gifted man like Kapteyn
should not be concerned with things that others could look after equally well.
Finally, what is the value or importance of a frequency curve in relation to the
structure of the Universe? I am happy that you admit there is an evil spirit
involved. Some form of exorcism is necessary and I would like to apply it.
In spite of this, Kapteyn kept himself busy with statistics, probably partly
in a sincere attempt to help his biological friends, but his stubborn desire to