Experimental Design - PowerPoint Presentation

Slide1

Experimental DesignSlide2

Experimental Design

AIM:

How

can we design and diagram a completely randomized experiment?

What is single and double blinding?

How can we design and diagram a randomized block design experiment

?Slide3

The Stat

Handout

Pine Trees Resisting Drought

Package Liners

Teaching MethodsSlide4

Tomato plant example

revisited

You wish to run an experiment to test a new fertilizer and the amount of sun that tomato plants get on the juiciness and taste of their tomatoes.

Factor – Sunlight

Full sun

Partly sunnyFactor – FertilizerFull Fertilizer DoseHalf Fertilizer DoseNo Fertilizer

How many treatments? 2 • 3 = 6 treatmentsSlide5

Name all six treatments

Full Sun – Full Dose Fertilizer

Full Sun – Half Dose Fertilizer

Full Sun – No Fertilizer

Partly Sunny – Full Dose Fertilizer

Partly Sunny – Half Dose FertilizerPartly Sunny – No FertilizerSlide6

Draw a diagram to depict this experiment –

assume we are only looking at the fertilizer factor

Assume there 24 plants from the garden store.Slide7

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment

.Slide8

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment.

Who should be blinded

?Slide9

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment.

Who should be blinded

?Slide10

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment.

Who should be blinded?

Subjects should not know which treatment they are receiving – placebo effect

.Slide11

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment.

Who should be blinded?

Subjects should not know which treatment they are receiving – placebo effect.

Experimenters themselves often subconsciously behave in ways that favor what they believe [expectations].Slide12

Blinding –

what is it and why do we need it?

Humans are very susceptible to errors in judgment.

Who should be blinded?

Subjects should not know which treatment they are receiving – placebo effect.

Experimenters themselves often subconsciously behave in ways that favor what they believe [expectations].

In general, anyone who could impact the outcome should be blinded.Slide13

Single vs. Double blinding

Two main classes of individuals who affect the outcome of the experiment

:Slide14

Single vs. Double blinding

Two main classes of individuals who affect the outcome of the experiment:

Those who could

influence

the results

SubjectsTreatment administrators

techniciansSlide15

Single vs. Double blinding

Two main classes of individuals who affect the outcome of the experiment:

Those who could influence the results

Subjects

Treatment administrators

technicians

Those who evaluate the resultsJudges

PhysiciansSlide16

Single vs. Double blinding

Two main classes of individuals who affect the outcome of the experiment:

Those who could influence the results

Subjects

Treatment administrators

technicians

Those who evaluate the resultsJudges

Physicians

If

everyone

in one of these classes is blinded –

Single Blind

.Slide17

Single vs. Double blinding

Two main classes of individuals who affect the outcome of the experiment:

Those who could influence the results

Subjects

Treatment administrators

technicians

Those who evaluate the resultsJudges

Physicians

If everyone in one of these classes is blinded – Single Blind

.

If

everyone

on both of these classes is blinded –

Double Blind.Slide18

Back to our tomato example

Tomatoes are not people. Do we need blinding?Slide19

Back to our tomato example

Tomatoes are not people. Do we need blinding?

Double blinding should be used for most reliable results:

Blind the caretakers of the tomato plants

They could influence the results.

Blind the judges of the juiciness / tastiness

They evaluate the results.Slide20

Mediation for anxiety example

An experiment that claimed to show that mediation lowers anxiety proceeded as follows. The experimenter interviewed the subjects and rated their level of anxiety. Then the subjects were randomly assigned to two groups. The experimenter taught one group how to meditate and they meditated daily for a month. The other groups was simply told to relax more. At the end of the month, the experimenter interviewed al the subjects again and rated their anxiety level. The meditation group now had less anxiety. Psychologists said that the results were suspect because the ratings were not blind. Explain what this means and how lack of blindness could bias the results.Slide21

Mediation for anxiety example - solution

Because the experimenter knew which subjects had learned the meditation techniques, he (or she) may have had some expectations about the outcome of the experiment: if the experimenter believed that meditation was beneficial, he/she may subconsciously rate that group as being less anxious.Slide22

Pain relief study example

Fizz Laboratories has developed a new pain-relief medication. Sixty patients suffering from arthritis and needing pain relief are available. Each patient will be treated and asked an hour later, “About what percentage of pain relief did you experience?”

Why should Fizz not simply administer the new drug and record the patient responses?

Outline the design of an experiment to compare the drug’s effectiveness with that of aspirin and of a placebo.

Should patients be told which drug they are receiving? How would this knowledge affect their reactions?

If patients are not told which treatment they are receiving, the experiment is single blind. Should this experiment be double-blind also? Explain.Slide23

Fizz labs solution

If only the new drug is administered, there is nothing to compare to.

Randomly assign 20 patients to each of the treatment groups. After treatment, ask about pain relief. Compare average pain relief experienced in each group.

The subjects should not know which drug they are receiving – expectations impacted.

Yes, if not double blind, the researchers may subtly influence subjects into giving responses that support the new medication.Slide24

Now let’s talk about blocking

Let’s return to our tomato example, suppose we know that our tomato plants were purchased from two different nurseries (A and B) and that the conditions in the two nurseries are not the same

.Slide25

Now let’s talk about blocking

Let’s return to our tomato example, suppose we know that our tomato plants were purchased from two different nurseries (A and B) and that the conditions in the two nurseries are not the same.

Since the stores may vary in the care they give their plants – this may impact their juiciness and tastiness

.Slide26

Now let’s talk about blocking

Let’s return to our tomato example, suppose we know that our tomato plants were purchased from two different nurseries (A and B) and that the conditions in the two nurseries are not the same.

Since the stores may vary in the care they give their plants – this may impact their juiciness and tastiness.

We are better off BLOCKING by store

.Slide27

Now let’s talk about blocking

Let’s return to our tomato example, suppose we know that our tomato plants were purchased from two different nurseries (A and B) and that the conditions in the two nurseries are not the same.

Since the stores may vary in the care they give their plants – this may impact their juiciness and tastiness.

We are better off BLOCKING by store.

BLOCKING isolates the effect of the some shared characteristic that may impact the response variable.Slide28

Blocking tomatoes

We assign the plants to blocks based on store. This part is NOT random.

Then, we assign experimental units to treatments randomly WITHIN each block.

Notice that the

response variable

is compared ONLY

within blocks.Slide29

Statistically significant

So, we got a difference. How do we know if this difference is large enough to matter?Slide30

Statistically significant

So, we got a difference. How do we know if this difference is large enough to matter?

A result is statistically significant if it is unlikely to happen by random chance alone.