RE: Diabetes, Fasting Glucose Analysis, And How Likely Am I To Die Anyway?
I was about to say that for answering your first two questions, it would be necessary to get the error bars, that may also help us to understand which proportion of the participants are filling each age category.
And to answer the last one, over the young vs old hazard ratio, we must be careful of the fact that the error bars for the young guys are way higher. Before developing my thought, I am wondering whether the band corresponds to a 1sigma or 2sigma spread. In the former case, the would be discrepancy is also kind of compatible with a no discrepancy at all, the differences between both being below 2 sigmas.
If we forget about the error bar, shouldn't we take figure 1 into account to get a clue?
PS: it was nice to read from you after such a long time! Your divulgation skills are actually way better than mine :)
Your point for the error in the second figure is good to keep in consideration. How confident can we really be that there is a statistically relevant difference between those two age groupings. There is clearly a huge variance in the data for the younger group.
That gets at another point, confidence in conclusions we could try to draw from the data. It's defined by the quality of the data.
However my question was more in identifying why younger people might have a higher hazard ratio then older people. The answer I am thinking of is very certainly not the only one, nor the only thing worth considering from the data. I was curious what people might come up with.
Let's see. Ignoring the error, I would tend to say this may be connected to the first figure and the genuine differences between the two age categories (that are not really visible on the third figure). But I may be wrong here :)