> PS this story doesn't make sense because you can tell when a voltmeter is out of range.

It's also a bit sloppy with the bounds. To me, "reads only as far as 100" includes 100V, whereas the measured voltages only go as high as 99V.

While we're complaining, you can't measure the "voltage" of an electron tube out of circuit.

Sure you can, just buy a SMU (source measurement unit, few multimeters cobbled together with 4 quadrant power supply)

I think the point is that the sample readings were in the range of voltmeter, but if some sample crossed 100V it would show as 100V in the voltmeter. So even though no sample crossed 100V, the data is clipped still.

That's not good statistics - just because the precision of measurements falls off sharply at 100V doesn't mean that out-of-range readings aren't valid measurements.

Assume we got one out of range measurement out of 100. We can't say anything about variance of reading if we don't have any prior on variance ie using bayesian statistics. That's the joke that traditional statistician can't figure out the obvious thing.