SOC222H5 Lecture Notes - Lecture 7: Chi-Squared Distribution
Document Summary
How do we know that there is some type of association based on a bivariate table: when comparing the column percentages of the cells, they differ from their respective row percentages. Consider the counter example: when absolutely no relationship exists between the 2 variables. The cell percentages should be equal to their respective row percentages. We expect that the percentage in each cell (representing the % of those with two particular attributes) e. g. conservative and agree and not conservative and agteed is no different that the percentage . Yes the cell column percentages suggest an association between political party identification and agreement with trump. Are these percentage differences large enough to say that one attribute. That test is called a chi-squared test. Compares the observed frequency against the expected frequency. By doing the formula the product remains the same no matter which cell you do first. Then take each product and divide it by the same expected value.
