Applied Data Analysis
Assignment 3
ANOVA (oneway and factorial) and t-test
Date: 26-03-2021
, Excersise 1: pretest – t-tests
a. Check assumptions
For education:
1. Independence of observations
We cannot check for this in SPSS, independence of observations is a design matter. Based
on SPSS data we don’t know whether person 1 is connected to person 2 for instance.
2. Normality
To check whether the populations are normally distributed we can first check visually
using histograms. We don’t see a lot of deviations from normality, so graphically we can
assume that the variable life satisfaction might be normally distributed in both
populations.
The test of normality (kolmogorov-smirvov) is non-significant for both high and low
education, which means we maintain the null hypothesis and assume our data is normally
distributed. Dlow(91) = .059, p = .811 ; Dhigh(92) = .062, p = .831
Numerically, the standardized skewness for the low education group is -.233/.253 = -0.92.
Standardized kurtosis of the residuals of low education is -.194/.500 = -0.39. Both values
are lower than 3, which means that skewness and kurtosis should not be a problem.
The standardized skewness and standardized kurtosis for high education are also lower
than 3. The standardized skewness is -.118/-.251 = -0.47. Standardized kurtosis is
Assignment 3
ANOVA (oneway and factorial) and t-test
Date: 26-03-2021
, Excersise 1: pretest – t-tests
a. Check assumptions
For education:
1. Independence of observations
We cannot check for this in SPSS, independence of observations is a design matter. Based
on SPSS data we don’t know whether person 1 is connected to person 2 for instance.
2. Normality
To check whether the populations are normally distributed we can first check visually
using histograms. We don’t see a lot of deviations from normality, so graphically we can
assume that the variable life satisfaction might be normally distributed in both
populations.
The test of normality (kolmogorov-smirvov) is non-significant for both high and low
education, which means we maintain the null hypothesis and assume our data is normally
distributed. Dlow(91) = .059, p = .811 ; Dhigh(92) = .062, p = .831
Numerically, the standardized skewness for the low education group is -.233/.253 = -0.92.
Standardized kurtosis of the residuals of low education is -.194/.500 = -0.39. Both values
are lower than 3, which means that skewness and kurtosis should not be a problem.
The standardized skewness and standardized kurtosis for high education are also lower
than 3. The standardized skewness is -.118/-.251 = -0.47. Standardized kurtosis is
