18 - The outcomes: children and mothers

Summary

Stephen Hillier: This distribution of age of onset of menopause — you gave a median but there were two populations. Is that statistically significant?

Gita Mishra: Yes. The actual distribution you saw is skewed. That's the observed distribution and it is definitely not a normal distribution. So, we think that distribution fits these data well. It turns out that this skewed distribution is a mixture of two distributions; one distribution is normal with an early age of menopause around 40, and the other one is normal with a mean age around 50. If you simulate these data with normal distribution with these two and you mix it, a third and two-thirds, you get a distribution that fits the observed distribution perfectly.

Stephen Hillier: Perhaps this is a naïve question, but have you tried to relate any of the other factors independently to the two groups to see whether there is any difference?

Gita Mishra: Yes. We found a lot of early factors determine you to be in this [earlier] distribution and there was an interaction. For instance, the mother's age of menopause: if the mother had a very early age menopause the person was more likely to be in the earlier distribution. Weight at 2 years: if a child was lighter at 2 years old then she is more likely to be in the first distribution. And the other thing that was quite surprising was the effect of parental divorce.