How sure are we?

Every curve on this site is an estimate. This page shows, marker by marker, how much data stands behind it and how much its shape depends on the method.

The mortality curves

For each marker: people and deaths behind the curve, how far the confidence band spreads at the ends of the data (wider is less certain), the evidence that the curve bends, and whether a more flexible fit changes the picture. Sort by clicking a column heading.

The drift results

Which HRS measures could be corrected for measurement noise. Without a correction, drift toward the middle cannot be told apart from noise, so nothing is claimed for those measures.

The same markers and death in China

Hazard ratio per standard deviation of each marker measured at the first CHARLS visit, adjusted for age and sex, beside the NHANES estimate. Each is per its own population's standard deviation, which differ, so compare direction and rough size rather than exact values. CHARLS deaths are family-reported in exit interviews, with no registry linkage.