Tables and Non Parametric Tests Lecture 5 Nonnormal

Tables and Non Parametric Tests Lecture 5

Non-normal Data Binomial data Log-normal data Transform Data Compare the means of the transformed (normal) data Really non-normal data

Binomial Data Are the proportions of Turks in Aalborg and Århus the same? Non. Turks Aalborg 465 Århus 358 Turks 35 42

Are the proportions significantly different? Non. Turks Aalborg 465 Århus 358 Turks 35 7. 0% 42 10. 5% Compare 3. 5% (10. 5 – 7. 0%) with suitable SE.

Another Approach Observed Expected Non. Turks Aalborg 465 35 Århus 358 42 In total 77 turks in a 900 sample, i. e. 8. 6% Non. Turks Aalborg 457 43 Århus 366 34 We expect 34 turks in Århus (8. 6% of 400)

Same proprotion in Aalborg and Århus? Observed Expected Non. Turks Aalborg 465 35 Århus 358 42 Observed and expected should be close Non. Turks Aalborg 457 43 Århus 366 34

How to do it in SPSS …or data could be organized in 900 rows

Cross-Tabs

Tricks

Output Expected values Proportions Test Statistic P-value

Binomial One-Sample Is proportion equal to 10% Two-Sample Proportions in Aalborg and Århus are equal 1. Calculate proportion and 95% CI 2. Is 10% in the CI? …or use SPSS as I will show later K-Sample Proportions in Aalborg, Randers, Vester Hjermislev and Århus are equal Cross-Tabs handles two or more cities (categories)

One-Sample (Symmetry or Location) Kiama Blowhole Data • Highly skew distribution • Average approx 40 sec Normal distributed ? Median equal to 40 sec? • Rarely above 100 sec Only above 100 sec in 1% of the eruptions?

Location of median Median equal to 40 sec? Only above 100 sec in 1% of the eruptions?

Median equal to 40 sec?

Output NPar Tests Median equal to 40 sec? NOPE!

Only above 100 sec in 1% of the eruptions?