1

My question is about the Chi Square goodness-of-fit test. Not to be mistaken with the test for association or the test for correlation (both called Chi Square).

Let's say I have following variable 'education' (fictive example):
- high school: 10
- college: 4
- university: 3

Now we can know for instance that the distribution for the general population is:
- high school: 100
- college: 40
- university: 30

In this case I have a small sample size. If I run the Chi Square I might run into the problem of: '1 cell (50%) have expected frequencies less than 5'. In that case we cannot use the Chi Square. We can try to combine categories to overcome this problem, but that's not always possible. So I wonder, what alternative is there then? And how to perform it in SPSS?

Fico
  • 131
  • 4
  • 1
    Search our site: http://stats.stackexchange.com/search?q=Fisher%27s+exact+test. Some of these appear to answer your question fully, such as http://stats.stackexchange.com/questions/76669. Some additional ideas, as well as principles, are offered at http://stats.stackexchange.com/questions/135796. – whuber Jul 24 '16 at 16:43
  • A fisher test verifies if there is an association between 2 variables. But I only have one variable.. – Fico Jul 24 '16 at 18:16
  • That's right: but the alternative approaches and principles found in those threads apply. – whuber Jul 24 '16 at 20:27

0 Answers0