We all know that Chi-square is a statistical test used to observe the association between two categorical variables with 2 levels. If we have two categorical variables both of them have >=3 levels and the (33.3%) have expected count less than 5, so the result of the chi-squared test will not be accurate. (Then in those cases we have to use "Fisher's Exact test)
However, In statistics, G-tests are likelihood-ratio or maximum likelihood statistical significance tests that are now used in cases where chi-squared tests have been previously recommended. For samples of a reasonable size, the G-test and the chi-squared test will lead to the same conclusions. But the approximation to the theoretical chi-squared distribution for the G-test is better than for the Pearson's chi-squared test.
So, why are we using the chi-square test still now, while the g-test was recommended since 1981?
For two reasons I think:
1. lack of knowledge about these tests- new tests of often tested hypotheses take a while to filter through the research community. Though it may be surprising to some, Statistics journals are perhaps not exciting reading for some researchers and 2. it often doesn't matter much which test one chooses(see https://en.wikipedia.org/wiki/G-test ). Given these 2 things I think, it will take a while yet. Best, David Booth
Thanks, Dr. David Eugene Booth and Dr. Steve s.
I am planning to use G-test to show the association between categorical variables in my upcoming analysis instead of the chi-square test.
In the dataset, several variables contain some factors. Thus do not meet the requirement of the chi-square test (e.g., cell count
The rationale behind your selection is more important than your selection.
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity