home *** CD-ROM | disk | FTP | other *** search
-
- THE CHI-SQUARE TEST
- for 2X2 tables
-
-
- The Chi-square test applies to
-
- testing the independence of rows and
-
- columns in a 2X2 table. Chi-square is
-
- applied when certain conditions are
-
- satisfied. They are as follows:
-
-
- 1) No cell has an EXPECTED value less
- than five.
-
- 2) Never use percentages. Use actual
- counts or frequencies.
-
- 3) The samples (classifications) are
- independent. For example, if some
- of the men and women in the
- preceding article had been
- married, independence would have
- been violated, since a strong
- relationship already exists
- between husband and wife on
- political party.
-
-
- In the example of the previous
-
- section, we ask whether the proportion
-
- of males voting Democratic is the same
-
- as the proportion of females voting
-
- Democratic. Of course, we know that
-
- 60% (15/25) of the men in our sample
-
- will vote Democratic while 41% (7/17)
-
- of the females will vote Democratic.
-
- So much for the obvious. A statist-
-
- ical test of hypothesis helps to
-
- answer the more general question:
-
- IN THE ENTIRE COUNTRY, IS THE
-
- PROPORTION OF MALES VOTING DEMOCRATIC
-
- THE SAME AS THE PROPORTION OF FEMALES
-
- VOTING DEMOCRATIC?
-
- We know it is not true for our
-
- sample, but is that sufficient? Maybe
-
- we drew a weird sample.
-
- Chi-square is a statistic which
-
- helps to answer the question. It
-
- tells us how likely it is that we
-
- 'just drew a weird sample'.
-
- We deal numerically with 'likely' by
-
- calculating the probability of an
-
- event. The smaller the probability,
-
- the less likely the event. Thus
-
- if an event has a probability of .05
-
- it is expected to happen 1 time in 20.
-
- A probability of .01 means it is
-
- expected to happen 1 time in 100, and
-
- so on.
-
- The probability associated with a
-
- statistical test is referred to as the
-
- 'significance level of the test'.
-
- ----< continued in next article >-----
-