《研究方法》研究生教学课件.ppt

Statistical Hypothesis Testing Statistical hypothesis testing estimates whether a relationship (or difference) observed on the cases studied suggests there is a relationship (or difference) among the population from which the sample was drawn. Example: A sample correlation coefficient may be assessed to decide whether any relationship exist among cases in the population. Hypothesis Testing Logic and Procedures Statistical hypothesis testing sequence of steps Specify Statistical Hypotheses Hypothesis testing has two statistical hypotheses. Research hypothesis (H1) It represents an investigator’s expectation about what is true of the population parameter. Example: Salary level is positively related to years of work experience Null hypothesis (H0) The hypothesis of no association or no difference. Example: There is no relationship between the two or that the relationship is negative Research hypothesis : Salary level is positively related to years of work experience. research hypothesis, H1: βYX 0.0 null hypothesis, H0: βYX 0.0 The research hypothesis is directional because it anticipates a positive relationship between salary level and years of experience. A directional research hypothesis is called one-tailed. Specify Statistical Hypotheses Research hypothesis : The average age of women in a population differs from the average age of men. research hypothesis, H1: μw ≠ μM null hypothesis, H0: μw = μM The research hypothesis is nondirectional. It does not suggest that women are younger or older than men, only that the average age of the two groups differs. A nondirectional research hypothesis is called two-tailed. Specify Statistical Hypotheses Significance Level Statistical significance level : The risk they take of accepting the research hypothesis (H1) when the null hypothesis (H0) is true. Probability sampling involves uncertainty. The probability sample drawn from a population may not be rep