If the FBOE is due to earlier pregnancies changing the chemicals present in the mother’s womb during later pregnancies, then the absence of non-surviving children from the data could actually be a problem with the data.
No? If non-surviving children are absent from the data, then the birth position that we have for a cardinal (1st born, 2nd born, etc) is really a lower bound. A cardinal who our records show as a 2nd born might actually be a 3rd or 4th born, but cannot be a 1st born.
That’s true, but the probability mass function for total sibship size we estimate here would left-shifted if the non-survivors are absent. The reported family sizes would be smaller in kind.
I might be wrong that this would neutralize the effect you’re pointing out, but I think it does.
If we assume that the likelihood of a pregnancy leading to a child who lives long enough to be in our records is independent of the child’s order / mother’s age, then I agree that that the effects on the average birth order and average family size of our cardinals should in some sense cancel. However, it should still increase the error bars on both numbers and therefor the uncertainty of any conclusion. And I’m not sure I’d expect survival to be independent of order.
It doesn’t look to me like non-surviving children are reported in this data, so no.
However, the reported results doesn’t change when you just look at it century by century.
If the FBOE is due to earlier pregnancies changing the chemicals present in the mother’s womb during later pregnancies, then the absence of non-surviving children from the data could actually be a problem with the data.
Aren’t non-surviving children as likely to make a cardinal higher in pregnancy order as they are to make him lower in that order?
No? If non-surviving children are absent from the data, then the birth position that we have for a cardinal (1st born, 2nd born, etc) is really a lower bound. A cardinal who our records show as a 2nd born might actually be a 3rd or 4th born, but cannot be a 1st born.
That’s true, but the probability mass function for total sibship size we estimate here would left-shifted if the non-survivors are absent. The reported family sizes would be smaller in kind.
I might be wrong that this would neutralize the effect you’re pointing out, but I think it does.
If we assume that the likelihood of a pregnancy leading to a child who lives long enough to be in our records is independent of the child’s order / mother’s age, then I agree that that the effects on the average birth order and average family size of our cardinals should in some sense cancel. However, it should still increase the error bars on both numbers and therefor the uncertainty of any conclusion. And I’m not sure I’d expect survival to be independent of order.