yego.me
💡 Stop wasting time. Read Youtube instead of watch. Download Chrome Extension

Biased and unbiased estimators from sampling distributions examples


4m read
·Nov 11, 2024

Alejandro was curious if sample median was an unbiased estimator of population median. He placed ping-pong balls numbered from zero to 32 in a drum and mixed them well. Note that the median of the population is 16. He then took a random sample of five balls and calculated the median of the sample. He replaced the balls and repeated this process for a total of 50 trials.

His results are summarized in the dot plot below, where each dot represents the sample median from a sample of five balls. So this, right over here, is a sampling distribution of the sample median. You have a population of balls. We know a parameter; we know that the median of the population is 16. This, right over here, is the population.

Then, he keeps taking samples of five balls. So that is a five-ball sample right over there. He calculates the statistic, the median of the sample. He calculates the median there, then he does it again, he calculates the median, then he does it again, he calculates the median, and then he does this 50 times. He does this 50 times, and so what you see plotted right over here, you see 50 dots. Every time he took a sample of five balls and calculated the median, we see that is one of the dots here.

For example, four times he calculated a median of 20. Two times he calculated a median of 19. One time he calculated a median of four. Now they ask us, based on these results, does the sample median appear to be a biased or unbiased estimator of the population median? So pause this video and see if you can come up with an answer to that.

In order to be an unbiased estimator, and that is what you want your estimator to be, you want it to be unbiased. The sampling distribution for that statistic has to be evenly distributed about the true parameter for the population. For example, in this situation, the true parameter for the population—they say note that the median of the population is 16. So that's the thing that you're trying to estimate.

This is where the median of the population is. If we look at the sampling distribution, we see that it is roughly balanced to the left and to the right of that true parameter. So I would say that this is looking pretty unbiased because the sampling distribution is evenly balanced to the left and the right of the true median of the population, that true parameter.

If for some reason, the distribution looked something like this, then it would not be unbiased. Or if the distribution looked something like this, if the distribution if the distribution looked like this, if our sampling distribution was like this, this would mean that we're consistently underestimating the true parameter. And if our distribution looked like this, this means that our statistic is overestimating the true parameter.

Let's do another example. The dot plots below show an approximation to the sampling distribution for three different estimators of the same population parameter. If the actual value of the population parameter is 5, which dot plot displays the estimator with both low bias and low variability? Pause this video and see if you can come up with the answer.

So the population parameter that we're trying to estimate is five. So let's show that on each of these distributions, these approximations of the sampling distributions. So well, over here we have our scale, but this should be one, two, three, four, five, six, seven, and one, two, three, four, five, six, seven. And so our population parameter is five, so that is right over there; that is right over there, and this is right over here.

So which of these statistics seems to be biased? Let's just start there. Pause this video. Which of these look biased to you? Well, the one that looks clearly biased is statistic C. When you look at its sampling distribution, it is consistently to the left of our true parameter, so it's consistently underestimating our parameters. So this one right over here is biased.

Now statistic A and statistic B both look reasonably unbiased. Remember, these are approximations of the sampling distribution. If you want to get closer and closer to the sampling distribution, the true sampling distribution, you would keep taking samples and keep calculating the statistic and keep plotting them on this distribution right over here.

But for statistic A, it looks reasonably balanced to the left and the right. It's not perfectly balanced, but it's reasonable. Remember, these are approximations, and statistic B is also reasonably balanced to the left and the right. Now they don't just say which one has a low bias, but they also say low variability.

I would say both statistic A and statistic B have relatively low bias, but if we talk about low variability, that's how much spread do they have, especially around the true parameter. It looks like statistic A has much lower spread than statistic B right over here.

Statistic B, you've calculated statistics that are two away from the true parameter, while in statistic A, you've calculated a statistic of certain samples that are roughly half away from the true parameter. So the statistic that has both low bias and low variability, that would be statistic A.

More Articles

View All
POLAR OBSESSION 360 | National Geographic
Eleven years ago was my first trip to Antarctica. I came down here to do a story about the behavior of the leopard seal. My name is Paul Nicklin; it’s my job as a photojournalist to capture the importance and the fragility of this place and bring this bac…
Debating Finance Junkies | Ponzi Factor | V-Log 6
Hi, this is Todd. Thank you for joining me once again for my last and final V log of the year. First, I want to apologize for being absent for so long. The last one I did was on the SP500 from almost two months ago. Unfortunately, I’m not gonna do a conti…
Introducing: Khan Academy Kids!
Hi everyone, Sal here with my three-year-old son Azad, and we’re excited to announce the launch of Khan Academy Kids, which is designed to take students like Azad, ages two to five, to become lifelong learners. Hi friends, welcome to my room! Kids love t…
Calculating correlation coefficient r | AP Statistics | Khan Academy
What we’re going to do in this video is calculate by hand the correlation coefficient for a set of bivariate data. When I say bivariate, it’s just a fancy way of saying for each x data point, there is a corresponding y data point. Now, before I calculate…
Introduction to exponential decay
What we’re going to do in this video is quickly review exponential growth and then use that as our platform to introduce ourselves to exponential decay. So let’s review exponential growth. Let’s say we have something that… and I’ll do this on a table here…
Landing the 900: The moment Tony Hawk became a legend | Rewind the '90s
NARRATOR: In 1995, Tony Hawk wins gold in the first X Games and goes on to dominate the decade. In 1999 he cranks it up a notch, attempting a never before done trick, the 900. HAWK: A 900 is a two and a half spin rotation sort of summersault in the air, …