Everyone, I have compiled initial statistics based on 106 responses so far.

This gives the failure probability and the +/- 95% confidence intervals for that failure rate. Obviously, we need more data points to get more confidence, especially at higher mileages.

**Everyone help out by voting!**
Also, there will likely be a skew towards more failures given that many people with healthy water pumps may not check out the thread.

The plot gives the failure rate as a function of mileage. In the x-axis I list the total responses for each mileage bin. For those of you interested in the calculation details:

Total response in each bin = n_i

Failure probability for each bin (p_i): p_i = #dead_i / n_i

95% Confidence Interval (CI): CI_i = +/- 1.96*[ p_i*(1-p_i) / n_i ]^0.5