MATH 1050Y Lecture Notes - Lecture 4: Family-Wise Error Rate, Null Hypothesis, Analysis Of Variance
14 Apr 2018
School
Department
Course
Professor
Document Summary
Data for one-way anova: comparing numerical values across different groups, one possibility: two sample t-test. Multiple comparisons: compare all three groups, we need 3 different hypothesis , this is called multiple comparisions, because we are comparing multiple pairs of means, we are creating more opportunities to mistakenly reject the null hypothesis. The more tests we do, the greater the probability that (cid:449)e"ll mistakenly reject the null hypothesis at least once. Increase in the familywise error rate: anova allows us to compare many groups simultaneously without increasing the familywise error rate. Hypotheses of anova: are they all the same or at lest one difference, h0: there is no relationship between the categorical variable and the numerical variabl e. Ha: there is a relationship; the population means of the categories differ. The conditions for anova: random sample: between groups and within groups, observations within groups are independent. Can loose requirement the method works well unless the distribution are vary form normal.
