# Measuring acceleration due to gravity in the lab

I am measuring the acceleration due to gravity in the lab with an electromagnet apparatus.

My textbook says to take the average time for a number of falls (keeping the height constant of course).

But I recall many moons ago being told to take the shortest value, not the average.

The explanation being that the electromagnet can retain some magnetism for a short time after being switched off. So the ball can take longer than it should but not shorter.

Any thoughts on this?

Should I go with the shortest value or the average?

• These are the things that I never understood in error treatments. And I am even a chemist :( Another I hate is measuring a rod with the same ruler. But I think you can be wise in this case . However there won't be space for much error analysis..... – Alchimista Sep 21 at 10:49
• @Alchimista What do you mean by '' measuring a rod with the same ruler '' ? – Kantura Sep 21 at 10:51
• Lab exercise in which I was asked to do error analysis doing exactly that :( which by the way isn't in principle different from your current task. Except that you likely got different time readings. – Alchimista Sep 21 at 10:53
• @Alchimista Was it the same person doing the measurement over and over ? Or were there different people involved ? – Kantura Sep 21 at 10:57
• It does not change much assuming decent operators. Especially not in the rod case. But note that I was commenting, not answering. I do have a fundamental trouble on these topic. Namely I can see the sense of averaging samples but I fail at the same for the same sample repeatedly measured, at least for measurements with no noise. I personally would be wise in your case. But let's wait for an answer. – Alchimista Sep 21 at 11:03

Random errors show up in your measurements because they are random. That is, when you measure the same thing many times you get results that are scattered. We generally assume the errors follow a normal distribution, so then we can calculate a standard deviation $$\sigma$$, and the final standard error from doing $$N$$ measurements is $$\sigma/N$$.