Pearson chi-squared test (χ2): differences?

  • Thread starter Thread starter Drudge
  • Start date Start date
  • Tags Tags
    Chi-squared Test
AI Thread Summary
There are two main types of Pearson chi-squared tests: the "test for fit of a distribution" and the "test of independence." The choice between these tests is crucial, especially in exam scenarios, as they differ in how theoretical values and degrees of freedom are calculated. Using the wrong test can lead to incorrect conclusions about whether a sample represents a theoretical distribution. The impact of this mistake largely depends on the specific theoretical distribution involved in the problem. To accurately assess the difference, one must specify the distribution in question.
Drudge
Messages
30
Reaction score
0
So, as far as I know, there are two χ2-tests: "test for fit of a distribution" & "Test of independence"

How big of a mistake is it to use the one instead of the other in an exam for example (of course all exams are all different to some degree, but generally)?

The only differences I can really find out is how each test counts the theoretical value(s) and the way in which the degrees of freedom are counted

For example a problem might be as follows:
a random sample from population X is, as a function of age, distributed as follows

10-20
5
21-30
4
31-40
3
40-41
9

And the equivalent theoretical values are: 6, 5, 4, 5

Question:

"Does the sample represent the theoretical distribution?"

So, you would use a "test for fit of distribution", but how much of a difference is it to use a "test of independence"?
 
Physics news on Phys.org
Drudge said:
So, you would use a "test for fit of distribution", but how much of a difference is it to use a "test of independence"?

That would depend on what theoretical distribution was stated in the problem.
 
Stephen Tashi said:
That would depend on what theoretical distribution was stated in the problem.

Non normal distribution.
 
"Non-normal" is not a specific distribution. If you really want to know "how much" difference it would make you must be specific about the distribution - and if you are specific then you can calculate the difference yourself.
 
Back
Top