Log in

No account? Create an account

Previous Entry | Next Entry

Follow-up post #1: del_c's graph

After my earlier musings on the lifespans of the women who have been married to the presidents and vice-presidents of the USA, del_c has had a go at doing a better graphic representation of the data. This is what he came up with by plotting age at death against birthdate for the 83 women in the sample, using a ten-point moving median, presented as a stepped line.

The big jump in longevity at the end of the nineteenth century is still there - accentuated, to be sure, by the very short life of Alice Roosevelt, but even without her it would be pretty clear.

I think that there is still a discernible drop in longevity among the women born in the earlier part of the nineteenth century compared with those born in the eighteenth. del_c is more sceptical. You can decide for yourselves.


( 2 comments — Leave a comment )
Aug. 15th, 2007 09:54 am (UTC)
Means versus medians
That moving median is sampling ten points ahead, as your ten-point average did, so the leap upward in the median age at death from 65 to 83 begins just as Alice Roosevelt enters the sample, not just as she leaves it.

That said, removing her—or changing the year she died for one 30 or 60 years later—doesn't change the median much at all, which is part of the beauty of medians for me: they're not as sensitive to the actual values of the extremes of the sample, because they're only looking at the rank. As long as Alice is in the bottom five of ten, how far down she is doesn't make a difference. If she is moved from the bottom to the top, all the median does is reads the next rank down from what it otherwise would have been.

(There is a little mean involved in the calculation, because I chose ten to keep to your scheme, so the median falls between two women, who must then be averaged. I would prefer to have sampled nine or thirteen points (4n+1 generally), so that all those steps would have been identifiable to the eye as lining up with an actual person, instead of the average of two people.)

I don't know why medians are neglected in popular statistics; they usually answer our questions about life and chance much more germanely than means do. I particularly get annoyed by how means are resorted to in discussions of national wealth and poverty, as in the joke about how Bill Gates walked into a bar and made everyone else a millionaire, on average. He doesn't do much at all to the median, and that fits my intuition of the bar's clientele much better.
Aug. 15th, 2007 09:38 pm (UTC)
Well, one would expect female life expectancy to drop in the early nineteenth century because, generally, it did. The medicalisation of childbirth was a bit of a disaster for women. Male doctors not washing their hands spread peuerperal fever (Sp?). Heard a great programme about the chap who discovered this: he wasn't listened to partially because he was a social infererior to many of the other surgeons, but also *becausse* he discovered that mortality rates were lower in the mid-wife run wards.

Other factors are that urbanisation proved poor for maternal health, and the switch from wet nursing to nursing bottles and also to formula foods were bad for babies who then passed infections around.

All of which is only faintly relevant to your topic of course.
( 2 comments — Leave a comment )

Latest Month

May 2019


Powered by LiveJournal.com
Designed by yoksel