Summary Statistics for Premasters

Short important summary

 Null hypothesis (H0): There’s no effect. This is the one we try to reject
 Alternative hypothesis (H1): There is an effect (women are more likely to wear a skirt
than men). If we can reject H0, this is the one that is SUPPORTED by the data, but
not PROVEN
 NHST = Null Hypothesis Significance Testing
 Level of measurement:
o Categorical
 Binary or dichotomous variable
 Nominal variable
 Ordinal variable
o Continuous
 Interval
 Ratio
 Always report de Standard Deviation (SD = 3.45) if you report the mean (M = 19.22)
 If you’re asked to calculate the mode and the mean you can also make a histogram
that visualizes it
 Making graphs: When describing a graph (histogram) mention the following:
- Difference between two variables (more/less than)
- CI’s overlap yes or no
- Conclude if statistical difference yes or no
- Always mention that you still need to perform a statistical test to determine the
exact p-value to determine statistical significance
 Outliers:
o Observed values that are ‘extreme’ (they don’t represent reality well)
o They may lead to a biased estimate of the mean, particularly when sample sizes
are small.
o The SD gives you an indication of whether there are outliers in the sample
 Syntax is done with ‘paste’
 To only select one case of a variable go to Data > Select Cases > and choose the
option to only get the data you’re interested in
 If they ask to be criticical on the questions, look at the question if the question is
asked clear enough (month/year for example)
 Simple scatter  Line can be included after the output is generated. Double click the
scatter > elements > fit line at total > linear > no “attach label to line”
 Actually always do recode into different variables instead of recode into same
variables
 There are different variables that can be computed with compute variable; means,
differences and sums
 A normal distribution (bell-shaped) has a Mean = 0 and a SD = 1
 Testing assumptions
1) Normality
2) Homogeneity of variances
3) Dependent variable must be interval or ratio
4) Data from different participants need to be independent
 Homogeneity: Levene’s test 
o Tests if variances in different groups are the same.
o Significant = Variances not equal
o Non-Significant = Variances are equal

, If highter than .05 (5%) then there’s no significant difference, the homogeneity of
variances is met
 Normality: Calculate the z-scores for Skewness/kurtosis  normal if:
o 95% of z-scores lie between −1.96 and 1.96.
o 99% of z-scores lie between −2.58 and 2.58,
o 99.9% of them lie between −3.29 and 3.29.
- Perform Kolmogorov-Smirnov test  Explore > Plots > Normality plots with tests
> Transformed  Normal when higer than .05
 To visualize the normality in a graph  perform Q-Q plot
 Check reliability (Cronbach’s alpha)  Analyze > Scale > reliability analysis > put
them all into items > statistics > check: item, scale and scale if item deleted >
continue. Output between .7 and .8 = good alpha. Report like: The mean of the scale
was 2.67 (SD = 0.59) and the reliability of the scale was good, α = .85.
Which test to choose for? And how?!!
 The T-test is commonly used to examine whether the means of two samples of
scores differ significantly from each other, or whether the mean of a single sample
of scores differs significantly from some pre-established scores.
o One-sample T-test (used when you want to know if the mean of one sample
differs significantly from some specified value “Is the group on average older
than 24?”)
o Independent-samples T-test (used when the scores are of the groups are
unrelated or independent; “Do boys drink more alcohol than girls?")
o Paired-Samples T-test or dependent T-test (used when scores are measured
within members of the same group, so scores are related; “prior/after”)
 One-way ANOVA: allows you to determine the significance of the differences among
3 (or more!) groups, rather than just 2.
 Follow-up tests:
o Post hoc analysis: You compare all means
o Planned contrast analysis: You do a specific analysis
 For each test there’s a different effect size to calculate:
o ANOVA: (Partial) Eta-Squared η2
o Post Hoc: Effect size (cohen’s d)
o Planned contrast: Pearson’s r
 Welch  One-way ANOVA: The Welch-test results are the ones you report if the
Levene’s test have indicated trouble with homogeneity.
 Two-way or factorial ANOVA: allows you to analyze the simultaneous effects of two
(or more) independent variables or factors on some dependent variable.
 Interaction effect (two-way or factorial ANOVA)
o Show how the effects of one IV might depend on the effects of another
o Are often more interesting than main effects
 Simple effects: with the main analysis you can check whether there IS an interaction
effect  Done via SYNTAX
 Correlation test
o Correlation analysis: “relationship or assocation”
o Examples correlation tests
 Regression analysis
o Regression analysis: if the question contains “predict”
Regression model is good if the individual outcomes deviade a lot from the mean and
deviate low from the regression line
 Steps for doing a regression analysis
 Chi-square

,o Df  1 means 2 rows, 2 col, DF  2 means 2 rows, 3 col

, Templates for all tests

 One-sample T-Test

Analyze > Compare Means > One-Sample T Test.

Cohen’s d

 Independent samples T-Test

Analyze > Compare Means > Independent-Samples T Test.

Cohen’s d

 Paired-samples / dependent T-Test

Analyze >Compare Means >Paired Samples T test.

 One-way ANOVA

Analyze > Compare means > One-way ANOVA: options > Descriptive, homogeneity
of variance test and Welch > Continue

Planned contrasts: Pearson’s r
Post hoc: Cohen’s d

 Two-way/factorial ANOVA

Analyze > General Linear Model > Univariate

Simple effects (SPSS Syntax)

 Correlation analysis

Analyze > Correlate > Bivariate

 Regression analysis

Analyze > Regression > Linear

 Chi-square test (crosstab)

Analyze > Descriptive statistics > Cross tab

Odds ratio