4 ms·
Does a variance 5 microseconds with a min of 38 milliseconds and a max of 47 milliseconds with 40 samples seem like a valid statistic?
by gmm1990 2y ago
Does a variance 5 microseconds with a min of 38 milliseconds and a max of 47 milliseconds with 40 samples seem like a valid statistic?
- menaerus 2y agoIn general, 40 samples would certainly not be enough for the experiment result to have statistical significance but perhaps in this case when the code being tested is very tight, e.g. just several tens of instructions, it may be enough. Code in this example is testing the CPU compute capabilities and the only source of variance is the CPU internals itself.