I am conducting a research study consisting content analysis of newspapers. I have three newspapers and two categories, and i want to find difference in the categories coverage.What type of chi square i can use?
Chi Square for association is used to see the association between two variables, for example, if you want to know if gender is associated with the type of paper one reads, then it should be Chi Square for association. However, this does not allow you to gauge the strength of the association, so sometimes, it is good to do Chi Square and Cramer's V for association.
Chi-square will tell you whether the distribution of results across your three newspapers is random. You might think of it this way: If you collapsed across those three categories to get a total percentage distribution on the other two categories, would you find that same overall distribution within each of the three newspapers, or would at least one of them have a different pattern?
This is about as far as you can go if your three newspapers amount to a nominal variable, where there is no ordering in them as categories. If, instead, the three represent an ordinal variable (e.g., low to high on some characteristic), then you should start with a Chi-sq to see if there is a non-random difference across the three. If so, the movie on to non-parametic statistics that would (presumably) treat the newspapers as a in independent variable that had an effect on the distribution of a dependent variable.
Dear David L Morgan, The three newspapers are nominal level measurement, and the two categories media freedom and accountability are also nominal in the form of categories. I want to measure whether there is a significant difference in media freedom and accountability coverage?
It sounds like you have a three category independent variable and two different dependent variables. It would help if each of your dependent variables were scored 1/0 so that you would be dealing with the percentage of the coverage in each paper that could be coded as either media freedom or accountability. Then the two tables that you would analyze would each be 3x2.
To clarify further with regard to your original question, Chi-square tests for independence between variables, i.e., whether or not there is a non-random relationship between them. If there is a a non-random relationship, then this is just another way of saying that the variables are "associated."