For a symmetric data set (One where negative and positive numbers are possible) the rules are:
In no case do three or fewer points trigger an out of control alarm until you get beyond 2σ.
Three points define a plane to a mathematician, or a line to an engineer, but don't quite suffice for a trend under Western Electric control chart rules.
That being said, I start to watch a process very closely when three consecutive points are on the same area...
How are negative minutes a possibility? Only positive are possible.
How are negative minutes a possibility? Only positive are possible.
Second time is coincidence.
Third time is enemy action.
..................... Auric Goldfinger
CA....
That is true if you assume you are starting with an unbiased coin/event etc. If you know there will likely be a bias in the outcome then its a safe bet you don't need more events.
Please use statistics correctly if you are you going to pretend to use them.
For example, there are not THREE measurements here. There were three debates. Each debate had a number of questions. The measurement to be used is the excess timing for each question.
Second, you have yet to express what the sigma for this event should be. If each candidate has 120 seconds to speak, the excess should be on the order of a few seconds. Accumulating a 180 plus overage during an entire debate is a lot of excess.
You have yet to specify what the sigma is you are using/claiming. I would assert, without doing the math, that a 3 minute overage would be outside the 3 sigma limit where a single occurance is sufficient to show bias.
You are assuming a normal distribution. That may not be a valid assumption for this process.