Statistics Sampling & Bias

Hi and welcome to www.educator.com.0000

Today we are going to be talking about sampling and bias.0002

The next couple of lectures they are going to be about the way that you collect the data and the different ways that are involved there.0006

It is really because we are going to be going into inferential statistics, so we are going to be talking about0020

the difference between descriptive statistics and inferential statistics.0026

In that introduction we are going to start off with how do you collect the data in order to do inferential statistics.0032

Because of that we are going to be talking about how these two methods are involved simply to collect data and experiments to collect data.0038

The book of this lecture is going to be based on sampling.0050

It is going to be about why we use sampling, the goal of sampling which is to avoid biased0053

and the types of biases that might occur because of different kinds of sampling.0058

First let us start off with the difference between the descriptive statistics and inferential statistics.0066

You may not have noticed so far what we have been doing is largely descriptive statistics.0072

Everything that we been doing so far like visualizing data, summarizing data, getting things0077

like the mean, median, and interquartile range, standard deviation.0083

All of these things are just about summarizing that data.0092

We have looked at doing that for one variable or two variables.0096

We have looked at getting mean, median, and mode and visualizing one variable at a time on bar graphs or histograms.0102

We also looked at things like regression and correlation and looking at scatterplot, so visualizing and summarizing two variables at the time.0110

Even though we can do all of that stuff so far, there are still lots of things we cannot do that are important.0126

Here is one example, so maybe we could do stuff like this.0134

A sample of 53 out of 100 voters will vote for candidate A but here is what we do not know, will candidate A win the election?0138

How likely is that?0148

We could summarize our sample, we can visualize it but we cannot make any conclusions about it for the actual population.0153

We cannot do generalization.0164

This is what skill or ability is called.0167

We cannot do generalization going from the sample to understand something about the population.0172

That is something we cannot do yet.0179

Here is an example of something that you cannot do.0183

A sample of 20 preschoolers who are raised with pets scored an average of 5.6 out of 10 on a biology class but a sample of those without pets scored 4.5.0185

Does this mean that having pets make children better in biology?0197

We have no idea.0201

I could tell you I could summarize for you the sample of 20 preschoolers with pets and without pets.0203

I could summarize that.0210

I could visualize that for you and I could do that with these two variables whether they have been raised with pets or not.0211

The pets are the variable which is categorical and their score on the biology test which is continuous.0219

I could do that but I still cannot tell you whether having pets actually cause children to be better at biology.0226

This is compare and find causation.0234

We cannot figure out causation so far and that is really going from the data in order to have a theory about causation.0245

This is something that we are also not able to do.0256

Number one thing we cannot do is we cannot do any generalization.0259

Number two thing we cannot do is we cannot say anything about one variable causing another variable, but causing changes in another variable.0264

First, in order to tackle generalization that is why we are going to take on sampling.0275

Sampling is going to help solve this problem.0279

In order to take on the problem of not being able to determine causation we are going to be using experiments to do that.0293

So, first, let us talk about sampling.0301

This is what a common generalization problem.0304

We have the sample that where 53 out of 100 voters will vote for candidate A but the question is will candidate A win the actual election?0309

Here our little sample is 100 potential voters.0319

If these 100 potential voters and we know what their opinion is like but the question is every voter in election,0327

we want to predict their behavior and that is the problem.0343

We only know about these 100 potential voters which is the subset and we do not know if they reflect every voter in the election.0347

And so what we are really looking for is a way to get a sample of this population that actually represents this population well.0365

Sampling is the method of choosing or the items or cases in your study.0369

And so that is the whole idea behind sampling.0394

It is a method of choosing your particular sample.0396

Here is the thing, in order to select a good sample and so what is a good example?0401

A good sample is one that is a representative of sample and a representative meaning that0412

this sample if it accurately reflects this population then that is a really good sample.0427

It is hard to guarantee a good example because there is just no way I could ever really know from looking at a tiny book if it actually reflects the population.0434

The only way I could really know that if I knew everything about the population then I can compare it to my samples.0445

Then all my samples are representatives but that is not the problem, we often cannot get the whole population.0450

How do I know if my sample is any good?0457

The bad news is we cannot really know but the good news is we determine a good sample by a good sampling method.0463

We are not going to judge your like recipe by how good the food actually turns out.0481

We are going to judge the recipe about by looking at how good the recipe or the techniques actually are.0488

In that way we do not judge a sample by actually knowing how representative it is but we think it is a good sample if it has a good sampling method.0495

And by good it, we mean unbiased.0505