PS296 Lecture Notes - Lecture 22: Null Hypothesis, Categorical Variable
Document Summary
Test for categorical data distributed across one dimension. Are the outcomes different enough from chance to conclude they are not random occurrences (or, different than what we expect) Involves a comparison of observed and expected frequencies. 2 has low power (ability to correctly reject the null hypothesis) when expected frequencies are small. We want at least 5 expected frequencies in each cell or the power is very low. Null: responding pattern is not different than what we expect by chance. Alternative: responding pattern is different than what we would expect (by chance) Subtract expected value from observed value, square it, and divide by expected value. Do this for each column, and sum all the columns together. We use frequencies, not percentages, in the chart. A chi-square goodness-of-fit test, at alpha = . 05, between hearing sound in the left and right ear produced a significant result, 2(1) = 5. 71, p < . 05.
