18.2 Purpose: determine the accuracy of your hypotheses due to the fact that you have collected a sample of data, not a census. Two approaches of hypothesis testing: = Classical statistics = Bayesian statistics > Extension ofthe classical approach > Consists of subjective probability estimates (degrees of belief) A ifference has statistical significance if there is good reason to believe that the difference does not represent random sampling fluctuations only. ‘Null hypothesis = Used for testing = Statement that no difference exists between the parameter and the statistic being compared toit ~The null hypothesis can neverbe proved, therefore we say it‘cannot be rejected’ instead of ‘accepted’ = We reject the null hypothesis when we find a statistically significant diference ~The probability ofa correct decision if the null hypothesis is true is 95 per cent. Altemative hypothesis, ~ Opposite of null hypothesis ~ There has been a change ‘Twortailed test ~ Non-directional ~ Exg. average could be more than 50 or less than 50 days, One-tailed test ~ Directional test = Eg. average could be more than 50 Hois true Hois true ‘Accept Ho Correct decision Type Werror (8) = Found not guity, = Guilty of cvime, unjustly Innocentof crime ccquitted ‘Accept Ha Type lerror Correct decision = Innocent, unosty = Guity,justy comited conicted nvhibie 182 One amd wot tet the Spr cnt evel of gic = hq annie ae @—— tc —ele a> 8 Power of the test: (1~ 8) -> 1~ chance of making a type il error How to reduce Type Il errors: = Shift the critical value closer to the 1, But then you also have to accept a larger a ~ Increase the sample size 18.3 STATISTICAL TESTING PROCEDURE Procedure for testing for statistical significance: 1. State the null hypothesis 2. Choose the statistical test > Criteria that can be used in choosing atest ‘© Power of the test (efficiency) © How the sampleis drawn ‘© The nature of the population © Type of measurement scale used 13. Select the desired level of significance > 0.05is the most commonly used, 0.01 as well > The exact level to choose is largely determined by how much «risk one is willing to accept and the effect that this choice has on & risk 4. Compute the calculated difference value 5. Obtain the critical test value 6. Interpret the test. For most tests ifthe calculated value is larger than the critical value, we reject the null hypothesis and conclude that the alternative hypothesis is supported, Probability (p) values ~The pvalue is compared to the significance level (a), and on this basis the null hypothesi either rejected or not rejected nP a: do not reject Ho 18.4 TEST OF SIGNIFICANCE Parametric tests: data are derived from interval and ratio measurements > more powerful 4 assumptions have to be met in order to use parametric tests: 1. Independence = The observations must be independent. That is, the selection of any one case should not affect the chances of any other case being included in the sample. 2. Normally distributed = The observations should be drawn from normally distributed populations. 3. Equal variances (homogeneity) ‘Measurement scales should be at least interval ‘A normal probability plot compares the observed values with those expected from a normal distribution. Ifthe data display the characteristics of normality, the point will fall within a narrow band along a straight line, Non-parametrictests: 1sed to tests hypotheses with nominal and ordinal data 3 questions to consider when choosing a particular significance test: 1. Does the test involve one sample, two samples or k samples? 2. If two samples or k samples are involved, are the individual cases independent or related? 3. Is the measurement scale nominal, ordinal, interval or ratio? ‘Additional questions: What isthe sample size? I there are several samples, are they of equal size? - Have the data been weighted? - Have the data been transformed? One-sample tests One-sample tests are used when we have a single sample and wish to test the hypothesis that it comes froma specified population. Parametric tests ‘The Z test or t-test is used to determine the statistical significance between a sample distribution ‘mean and a parameter. Use t-test for smaller sample sized (< 30). Non-parametric tests I the measurement scale is nominal, itis possible to use either the binomial test or the chi-square (0) one-sample test. Binomial: when the population is viewed as only two classes (male/female, successful/unsuccessful) and when sample size so small that X* cannot be used, 2