Lesson 11 - Testing for Differences and Normality in SPSS PDF

Preview

Summary

This is how to test for differences and normality in SPSS...

Description

Testing for Differences and Normality in SPSS - RC4111 Learning Objectives: 

Recap - what tests for differences allows us to do



Different types of tests



How to use normality in SPSS



Parametric tests in SPSS - T test and Paired T test



Non-parametric tests in SPSS - Mann Whitney and Wilcoxon

Testing for Differences Recap: 

Tests for differences investigate effects of one or more IV’s (with two or more levels) on a numerical (usually continuous) DV



The nature of the levels of the IV determine which tests to use (independent/related samples)

Independent and Related Samples: 

Samples - IV levels (groups)

Test for Differences (2 Groups): 

Independent samples o

Independent samples t-test 

o

Mann Whitney U test 



Non-parametric

Paired samples o

Paired samples t-test 

o

Parametric

Wilcoxon signed-rank test 



Parametric

Non-parametric

H0 - there is no difference between the measurements for the different groups

Outlook of Parametric Tests: 

Independent test o

Parametric - data needs to be normally distributed



o

Compares means of the group

o

Uses standard error of the mean (SE)

o

Independent samples

Paired T test o

Very similar to the independent samples t-test

o

Compares mean responses/measurements

o

Two related groups using standard error of the mean

o

Paired samples

Outlook for Non-Parametric Tests: 



Mann Whitney U o

Non-parametric data can be different from normal

o

Compares medians of the group

o

Uses ranks

o

Independent samples

Wilcoxon signed rank test o

Similar to Mann Whitney U test for independent samples

o

Compares median responses of two related groups]

o

Uses ranks (similar to interquartile range)

o

Paired samples

How do we decide on a statistical test? 1. Identify the dependent and dependent variable 2. What kind of data are these? 3. If applicable: is my data independent or paired?

4. If applicable: is my data normally distributed? Normality - how is my data distributed? 

To determine if data is normal or not, we look at histograms and perform tests for normality



2 common tests - Shapiro-Wilk and Kolmogorov-Smirnov



H0 - there is no difference between the distribution of the observed data and a normal distribution (the data is normally distributed)



Important - we want the null hypothesis to be true in order for our data to be normal (parametric)

Normality Tests - Generate P Values: 

P-Value >/= 0.05  accept H0; data is not different from normal



P-Value < 0.05  reject H0; data is different from normal



Why do we need to know this? o

To determine the appropriate statistical approach (which test to use)

Normality in SPSS: 

What does this mean - is the data normally distributed for both IV levels?



Only if all are normally distributed (p>0.05) can we use parametric statistics to test for differences/correlations etc. to answer our research questions

Normality in SPSS: 

Analyse  descriptive statistics  explore



Then select the dependent and independent variable (factor list), then in the plots option, select ‘normality plots with tests’. Click continue, then okay.

Example Data 1 - Effect of reproduction on the foraging behaviour of ewes. Energetically, producing milk is the most demanding period of the reproductive process for mammalian females. But do ewes spend more time eating when they are lactating in order to compensate for this demand. Is there a difference in the time ewes spend grazing when they do and do not have lambs? This would be related samples as you are looking at the same ewes but with lambs and without lambs. Data in SPSS (paired samples) 

Remember - each row is a case (ewe), each column is a variable (grazing time)

Example Data 2 - effect of sex on propensity to developing osteoporosis? According to literature, women are more susceptible than men; 1 in 3 over the age of 50 affected. But is this measurable before patients develop clinical signs of osteoporosis? Does sex affect bone density in healthy 50+ patients? This would be independent samples as the male’s results would not have an affect on the females results and vice versa.



Remember - each row is a case (patient), each column is a variable (sex and bone density)

Independent Samples in SPSS

Paired Samples in SPSS...