Psychology week 6 assignment 2
Week 6 Assignment Statistics Guide
Open data file (programeval.sav)
Before determining which statistical tests you choose (paired t-test vs. Wilcoxon signed- rank), you need to do three checks:
1. Missing Data
2. Outliers (via boxplots)
3. Normality (via histograms + Shapiro-Wilk test)
1.Checking for Missing Data in Jamovi
Paired tests require that each participant has both a pre score and a post score for your chosen outcome.
Steps
1. Click on Exploration → Descriptives
2. Move your selected pre and post-test variables (e.g., SDQ_pre and SDQ_post) into Variables.
3. In the right panel, check:
o “Missing data” (shows N valid and N missing)
o “Statistics” → “N”
What to look for? How many participants and how many missing for each variable.
Just report how many are missing for each variable. You do not need to do anything else as for paired analyses, Jamovi automatically drops cases missing either pre or post for that variable. This is called pairwise listwise deletion.
2. Checking for Outliers (Boxplots) in Jamovi
1. Go to Exploration → Descriptives
2. Add your pre and post variables
3. Select Plots → ✔ Boxplot
What to look for? Circles above/below the box indicate possible mild outliers. Outliers are suggestive of more extreme outliers. If you have extreme outliers, you should check skewness and kurtosis. If values are below 3, these are not extreme.
At this time, you should report any potential outliers but keep them in the dataset. This is so everyone is working with the same dataset.
3.Carry out Paired sample t-test – simplest way to assess for normality in Jamovi
1. Go to Analyses → T-Tests → Paired Samples T-Test
2. Click Pairs arrow button
3. Add your pre and post-test variables as a pair. For example:
o Left: SDQ_pre
o Right: SDQ_post
4. Under Assumption checks:
o ✔ “Normality test” (Jamovi repeats Shapiro-Wilk)
o ✔ “Q–Q plot”
What to look for? Are your difference scores approximately normal (Q–Q) and Shapiro p > .05? ➡ YES → Paired Samples t-test ➡ NO → Wilcoxon Signed-Rank Test
Also, be sure to check out histograms for your pre and post test scores (Analyses- Exploration- Descriptives – check off Histograms)
You can carry out paired sample t-tests if:
• The difference scores look roughly normal
• Shapiro–Wilk p > .05
• Outliers are mild/not excessive
If you decide to use paired sample test:
Under paired sample t-test, check:
o ✔ Descriptives
o ✔ Mean difference
o ✔ Confidence interval
o ✔ Effect size (Cohen’s d)
If you decide to use Wilcoxon Signed-Rank Test:
Use this when:
• Difference scores are non-normal
• Shapiro–Wilk p < .05
• Outliers present
• Ordinal or skewed measures
Steps
1. Go to Analyses → t-tests → Paired Samples
2. Keep in the variable pair (e.g., PHQA_pre → PHQA_post)
3. Check Wilcoxon Signed-Rank Test
4. Under options, select:
o ✔ Effect size (r)
o ✔ Descriptive statistics
You will get:
• Median pre
• Median post
• Median difference
• W statistic
• p-value
• Effect size r
Regardless of which test you run, create a Paired Means Plot
Steps
1. Go to Analyses → T-Tests → Paired Samples T-Test
2. After adding your pair, check:
o ✔ Descriptives Plots
Jamovi produces a line plot showing:
• Mean pre
• Mean post
• 95% CI error bars
Summary: Follow these steps
1. Check missing data
2. Check for outliers (boxplots)
3. Check normality (histogram + Q–Q + Shapiro)
4. Choose the test:
o ✔ Paired t-test if normal
o ✔ Wilcoxon if non-normal
5. Run the appropriate test
6. Create a graph (paired means or boxplot)
7. Report:
o Missing data
o Outliers
o Descriptive statistics
o Test statistic, degrees of freedom and p value
o Confidence intervals
o Effect size
o Interpretation
Template for Reporting
Descriptive statistics: Scores on [variable] at pre-test (M/Median = ___, SD/IQR = ___) and post-test (M/Median = ___, SD/IQR = ___) showed a(n) increase/decrease/no change.
Test result: A paired samples t-test / Wilcoxon Signed-Rank test indicated that this change was/was not statistically significant, t(df = ___) = ___, p = ___ / Z = ___, p = ___.”
Confidence interval: The estimated difference (post − pre) was , with a __% CI [___, ___]
Effect size: The effect size was Cohen’s d = ___ / r = ___, indicating a small/medium/large effect.
Interpretation: These results suggest that participants showed a(n) increase/decrease/no meaningful change in [variable] from pre- to post-test.