WEBVTT
00:00:00.012 --> 00:00:04.368
>> In general, samples underestimate the amount of variability in a population,
00:00:04.368 --> 00:00:08.881
because samples tend to be values in the middle of the population. Especially in
00:00:08.881 --> 00:00:13.314
a normal distribution, most of the values are centered here in the middle. So
00:00:13.314 --> 00:00:18.646
when we take samples from it, most of our values are going to be around here,
00:00:18.646 --> 00:00:24.038
since most of the values are in this area. Therefore the variability in this
00:00:24.038 --> 00:00:29.565
sample will be less than the variability of the entire population. To correct
00:00:29.565 --> 00:00:35.223
for this, we use something called Bessel's correction, where instead of dividing
00:00:35.223 --> 00:00:40.389
by n, we divide by n minus 1. Same within the variance. So what will dividing by
00:00:40.389 --> 00:00:45.413
n minus 1 do to the original standard deviation and variance? Will it make them
00:00:45.413 --> 00:00:47.843
bigger, or will it make them smaller?