taken from lecture given by dr. Iwan Ariawan Concept Statistical test is conducted to determine the observed difference/relation in sample is a real difference/relation in population or occured by chance only (due to sampling process) Result of statistical test: “probability to get the observed result as seen in the sample or more extreme if (in reality) there is no difference/relation in the population” The statement “no difference/relation in population” is known as null hypothesis Concept Statistical test is conducted to determine the observed difference/relation in sample is a real difference/relation in population or occured by chance only (due to sampling process) Result of statistical test: “probability to get the observed result as seen in the sample or more extreme if (in reality) there is no difference/relation in the population” The statement “no difference/relation in population” is known as null hypothesis Concept Statistical tests will show probability (p value) that the result we see in sample is in line with null hypothesis If p is large then result is in line with null hypothesis, or we say the null hypothesis is accepted On the other hand, if p is small then result is not in line with null hypothesis or null hypothesis is rejected We use a as the threshold for determining p value is small or large Concept If we reject H0 There is a possibility of type 1 error (a), we incorrectly concluded that there is a difference/relation in population (as observed in sample) but actually there is no difference/relation in population If we accept H0: There is a possibility of type 2 error (b), we incorrectly concluded that there is no difference/relation in population (as observed in sample) but actually there is difference/relation in population Difference/Relation (or stated as statitically significant) as concluded from statistical test IS NOT automatically means real substancial difference/relation Small differences/weak relation could be statistically significant in large sample The a and b Statistical Testing Procedure
Step 1: State null and alternate hypotheses
Step 2: Select a level of significance
Step 3: Identify the test statistic
Step 4: Formulate a decision rule
Step 5: Take a sample, arrive at a decision
Do not reject null Reject null and accept alternate
Choice of test statistics Choice of test statistics depend on: type of exposure/independent and outcome/dependent variables (categorical/continous) data distribution (normal/not normal) the need to adjust for confounder (simple/multi variate analysis) Simple Analysis Dependent/Outcome Variable