2-4.2. Paired Test in Excel

Loading...

En provenance du cours de University of Illinois at Urbana-Champaign

Inferential and Predictive Statistics for Business

223 notes

University of Illinois at Urbana-Champaign

Inferential and Predictive Statistics for Business

223 notes

Cours 6 sur 7 dans Specialization Managerial Economics and Business Analysis

This course provides an analytical framework to help you evaluate key problems in a structured fashion and will equip you with tools to better manage the uncertainties that pervade and complicate business processes. The course aim to cover statistical ideas that apply to managers. We will consider two basic themes: first, is recognizing and describing variations present in everything around us, and then modeling and making decisions in the presence of these variations. The fundamental concepts studied in this course will reappear in many other classes and business settings. Our focus will be on interpreting the meaning of the results in a business and managerial setting. While you will be introduced to some of the science of what is being taught, the focus will be on applying the methodologies. This will be accomplished through use of Excel and using data sets from many different disciplines, allowing you to see the use of statistics in very diverse settings. The course will focus not only on explaining these concepts but also understanding the meaning of the results obtained. Upon successful completion of this course, you will be able to: • Test for beliefs about a population.. • Compare differences between populations. • Use linear regression model for prediction. • Learn how to use Excel for statistical analysis. This course is part of the iMBA offered by the University of Illinois, a flexible, fully-accredited online MBA at an incredibly competitive price. For more information, please see the Resource page in this course and onlinemba.illinois.edu.

À partir de la leçon

Module 2: Statistical Inference Based on Two Samples

Does the medicine a person is taking to treat his condition really work better than a sugar pill? Is the new chip-enabled credit card more secure than the magnetic card? How do you know whether the claims being made about anything being “better than” or “faster than” a competitor are true? In this module we will learn to make this comparison.

2-4.1. Paired Tests9:14

2-4.2. Paired Test in Excel7:28

Rencontrer les enseignants

Fataneh Taghaboni-Dutta
Clinical Professor of Business Administration
University of Illinois, Urbana-Champaign College of Business Department of Business Administration

[SOUND]

To know whether or

not students will perform better at a standardized test after completing

a preparatory class, a group of 65 students are given a test before and

then again after having gone through the training.

So we want to test this claim at 5% level of significance.

This is an example of a pair test,

because the same people are being given a pre-test, then they go through a training,

and then they're given a post-test.

We're not comparing two different groups, one has been trained and

the other one has not been trained.

These are the same individuals.

This makes it a pair test, and pair tests,

by and large, are stronger tests when you're looking for differences.

So when you are doing this,

the hypothesis is just only about the mean difference, nothing else.

Now in this case, they're saying that the difference would be better.

So what they're saying is that alternative is that the average difference,

after going to test, through the test is going to be greater than zero,

which means that null is that, that difference is either zero or less.

So we're going to test this at 5% level of significance.

So let's see how would do this differently using Excel data analysis.

So here we are and these are the test scores

when the students went through pre-training and after training.

So each student has a score that is before they went through the training, and

after they have gone through training.

So we're going to test these values.

The ones in column B and column C.

So we go to data.

We go to data analysis just like before.

However this time we going to scroll down and

look for the t-Test Paired Two Samples for Means.

So not the one that has been using before.

So this is the one that we will use, the paired two sample for means, so click OK.

The rest of it is going to look very similar to what we were doing before.

Input for Variable 1 is in column B, this is before.

And this is the after test scores, Variable 2.

I'm going to mention that I have a label and then here I'm going to put a 0.

Now I could have also given you a value here to test.

I could have said the test scores goes up at least by ten points and

that's what you would have put here.

Alpha is 0.05. I'm going to Output Range here,

click here and then I just I'm going to put it somewhere here.

So here's our output.

And one of things that you notice right away is that right up

here it says that this is a t-Test for paired two sample.

And of course in a paired two sample,

we have to have exactly the same sample size that's a given,

because the same people are being tested in pre-test and after test.

So now we can say whether or not we have the same test scores or not.

So in our problem we had said that it would be greater than zero,

which means this was a one-tail test.

So we're going to focus on the one-tail test and that is right here.

And this value is extremely small.

Definitely smaller than .05.

So we will reject the null hypothesis, which means what?

Our null hypothesis was that the difference would be less than or 0.

We will accept the fact that based on our data,

the training seems to increase the test scores for these students.

Once again, we can use the information that we have here to even give

what is the confidence intervals for he differences in the mean.

So looking at the equation for the confidence interval, first of all,

I need to find out what is the difference that I see in the means right now.

And that's relatively simple.

I'm going to find the difference in the mean by taking Post-Training minus

Pre-Training, again it doesn't really make a difference but

I like to talk about how much we expect the test score to go up.

It would be easier to communicate here.

Then we have the t-value.

Now, we did a one-tail test and if I want to use that one-tail test,

and for that I used the t-value that I see here,

this is going to result in the 90% confidence interval.

Let me again remind you why that is.

For the one-tail test, the entire 5% is on one tail,

and this is a value of 1.669, which is what you see right here.

And we will have, the same on the other side because, again, confidence interval

is going to have a margin of error that's going to be added and subtracted from it.

So we can make a mistake on either end and there's a 5% chance on either side.

So if these are 5%, the confidence interval that you're coming up with

is going to be a 90% confidence interval.

So this t-value,

this t of alpha over two is going to be for 90% confidence interval.

Now then the equation asks for standard deviation of the differences,

and that part we cannot get from the output directly.

I'm going to have to calculate this and that's what I'm going to do.

So let's go back to our data.

Which is right here.

I'm going to create a column here called differences and

the differences are just going to be the simply for each student.

What was the difference between their score before and after?

So, it's the difference between Pre-Training minus Post-Training.

Again, it will not make a difference which one you subtract from which.

Because we are going to use this column for finding out the standard

deviation of the differences, so it doesn't make a difference.

So then, now that you have it for one, just put your cursor in the corner,

then you see the plus sign, double-click, and the entire thing will be picked up.

To find the standard deviation for the values that you see in this column,

column D is what we want, standard deviation of the differences.

Then it's simply =STDEV.S, which we have been using all along of the values you see

from here all the way to the 65th student, close the parentheses and press Return.

So the standard deviation of differences here is 111.58.

Now that I have this I can calculate the margin of error.

Margin of error is simply is equal to t-value times,

as you can see in the formula that you have, standard deviation of

differences divided by square root of sample size and

in this case you can pick either one of the ones that says 65.

That's all we need and

once I write all of it out I will find out that the margin of error is 23.1.

So now I can come up with a lower bound and upper bound of the concentrical.

So the lower bound is the mean difference minus margin of error and

the upper bound is the mean difference plus the margin of error.

But you see here is your 90% confidence interval of the differences.

So what do you expect?

You're 90% confident that the average score for

a student will go up after going to this training.

The average score for the student will go up by 24 points up to 70 points.

The true difference is somewhere between 24 and 70, so at the minimum,

a student will get a 24 point boost.

Which is not bad, and at the optimistic side it's 70.

But remember, any value between 24 and 70 is a plausible value.

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2019 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play