Topics for Today

Review of Midterm

Introduction to Hypothesis Testing


Hypothesis Testing

The Scientific Method

1.  The a priori identification of a ______of interest

2.  Accumulation of ______

3.  Determination of the hypothesis' ______

4.  ______the current state of knowledge

A Hypothesis is a testable conjecture


Hypotheses

Researchers define their ______as a statement of the conjecture they would like to prove.

Some examples of research hypotheses.

·  Treatment A works

·  The mean IQ in Group A is ______than the mean IQ in Group B

·  People in Vancouver have ____ disposable income than people in Toronto

·  Children attending a montessori elementary school are more ______to attend university


Null Hypothesis

The state of knowledge at the outset of an experiment can be considered the null state

A Null Hypothesis is an explicit statement of the current state of knowledge. (in other words – the hypothesis we wish to disprove through experimentation).


Examples of Research and corresponding Null Hypotheses:

·  Research: Treatment A works

·  Null: Treatment A does not work

·  Research: The mean IQ in Group A is higher than in Group B

·  Null: Group A & B are the same

·  Research: People in Vancouver have less disposable income than people in Toronto

·  Null: People in Vancouver & Toronto have the same disposable income

·  Research:Children who at a montessori elementary school are more likely to go to university

·  Null: type of elementary has nothing to do with getting into university

Science & Null hypotheses in a nutshell:

Accumulate information and determine if that evidence is sufficient to change the state of knowledge

.... collect data and analyze it!

Analogy to the Legal System

Legal System / Hypothesis Testing
Innocent until proven guilty / Assume we know nothing
Lawyers present evidence / Researchers run a study / collect data (sample)
Judge or Jury deliberates / Calculate a test statistic and p-value
Guilty or not Guilty / Accept or reject Null


Analogy continued

Juries, and judges, are not foolproof. Even if all evidence is complete and the lawyers did their job, it is ______that the _____ decision is made.

Defendant is Innocent / Defendant is Guilty
Not Guilty Verdict / Correct! / Mistake!
Guilty Verdict / Mistake! / Correct!

Similarly, even if we do everything right, because we took a sample of random variables, there’s a chance we could make an ______conclusion.

Null is True / Null is False
Accept Null / Correct! / Mistake!
Reject Null / Mistake! / Correct!

Because the null is the opposite of our research hypothesis, ______of the null is typically a ______result, while ______of the null is a ______result (we are convinced the research hypothesis is true).

Null is True / Null is False
Accept Null / Valid, negative result! / False Negative!
Reject Null / False Postive! / Valid positive!

Statisticians (inherently creative) assigned names to these types of mistakes

Null is True / Null is False
Accept Null / Valid, negative result! / Type II Error
Reject Null / Type I Error / Valid positive!


Hypothesis Test Example

Anemia in Jordan

Hemoglobin is a protein in red blood cells that carries oxygen from the lungs to body tissues. People with less than 12 grams of hemoglobin per deciliter of blood (g/dl) are anemic, and at risk for many dangerous complications.

A public health official in Jordan suspects that the ______level for all children in Jordan is ____ than 12 g/dl. She measures a random sample of 50 children.

[This is a hypothetical example constructed by me.]


Anemia in Jordan Cont’d

Individual: _____

Population: ______

Variable of interest: ______

Parameter of interest: ____ (μ)

Research Hypothesis: μ _ 12 g/dl

(a typical child in Jordan is anemic = average hemoglobin in Jordainian children is less than 12)

Null Hypothesis: μ _ 12 g/dl

(a typical child in Jordan is not anemic = average hemoglobin in Jordainian children is 12 or higher)


Anemia in Jordan Cont’d

Here is a reproduction of the earlier table with the possible decisions from this example:

Null is True / Null is False
Accept Null / claim there is no anemia / claim there is no anemia
Reject Null / claim there is anemia / claim there is anemia

So, this setting, here are the ______we can make

Type I Error: claim there is anemia when there really _____ anemia

Type II Error: claim there is no anemia when there really ___
Making decisions

Clearly we can make a wrong decision … just like a jury.

A jury makes their decision based on the arguments of the lawyers.

How do we decide which ______is supported by the ______?

Keep this in mind …

… like a jury who has to consider the defendant innocent until proven guilty …

… we must consider the ______, ____ until we’re ______otherwise.


Weighing the Evidence

Think back to the normal distribution.

Let’s say X is ______distributed, like someone’s height or weight.

What are ______for X? ______values?


Now … if we have a sample, and calculate the ______, whare are ______values of?

[ Remember, no matter what the distribution of X, the sample mean is normally distributed. ]


So, values of that are ______are very ______.

Key:

We don’t actually know what µ is!

But if we ______a value for µ, and is really ‘___’ away, from that ______value … then our ______value probably ______.

Revisit the Anemia in Jordan example:

The ____ hypothesis proposes a value of 12 mg/dl or more.

If the sample of the 50 children has mean of __ and ______of 2, it’s ______that the population mean is 12 mg/dl or more

What if the sample mean is _ with a standard deviation of 2? Is it ______the population mean is 12 mg/dl?

What if the sample mean is 5 with a standard deviation of __?

Today’s Topics

Hypothesis Testing

-  think of the jury analogy

-  can make mistakes

o  Type I error: reject null when true

o  Type II error: accept null when true

Reading

Start Chapter 7 … but stay with the notes (don’t’ go too far ahead)

Stat203 Page 2 of 18

Fall 2011 – Week 7 Lecture 3