Causal Claims

Lecture 14

June 17, 2007

 

I. What is a Causal Claim?

 

When you make a correlation claim, you are saying something about a population in the real world.

When you are making a causal claim, you are making predictions about potential future events. (hypothetical situations)

 

A. Various Uses of the Word Cause

The word cause can be used in many different ways.

1. THE cause = necessary condition

2. In order to be a cause, does not have to be a necessary or sufficient condition.

3. What we will mean by cause can also be called a causal factor.

 

 

Positive Causal Factor:  If everyone had condition X, there would be more people with effect Y than if no one had condition X.

Negative Causal Factor: If everyone had condition X, there would be less people with effect Y than if no one had condition X.

A causal claim is about hypothetical worlds, rather than the actual world.  We cannot go to those hypothetical worlds.  What can we do?  We can approximate these worlds, 'construct' them in the real world.

 

II. Evidence for Causal Claims

 

To look for causality:

A. Get two sample populations

1. The only relevant difference between the populations is that one population has the causal factor while the other does not.

2. We also want both samples to be representative of the population at large.

 

B. Check the Effect of the Rate in both Populations

1. If did A properly, would get evidence of what would be the case if causal factor changed.

 

C. Use Statistical Tools

1. Use same statistical tools as with correlations to see if the is a statistically large difference in effect in the two populations.

 

D. Who gets causal factor?

1. Experimental Group - gets the causal factor

2. Control group - does not have the causal factor

Both have to be representative of the population at large.

 

 

III. Three Things that can go Wrong in an Experiment

 

A. One of the groups is not representative of the population at large.

B. Other relevant differences between experimental group and control group besides alleged causal claim.

C. The results might not be statistically different.

 

 

IV. Experimental Design

 

A. Randomized Experimental Design

B. Prospective Experemental Design

C. ?

 

< Prev		Next >