Intermediate Psychological Statistics SPSS Moderation Analysis Sample (Taylor’s)
For information and reference only – not to be submitted as your own work.
This sample shows how the moderation analysis in a Taylor’s University Intermediate Psychological Statistics SPSS report could be written up in APA 7th style. It follows the parts of the Intermediate Psychological Statistics SPSS moderation analysis brief. It does not use the dataset you were given. All numbers below are illustrative output from a simulated sample of 200 adults, created only so that the statistics are internally consistent. Do not copy any value into your own report; run your own data and report your own results. For the planning advice behind this extract, read our guide on how to write an SPSS moderation analysis report.
Plan for the report
| Part | What to show | SPSS route |
|---|---|---|
| 1 Data preparation | Missing values, outliers, descriptives, correlations | Frequencies, Explore, Descriptives, Correlate |
| 2 Assumptions | Linearity, normality of residuals, homoscedasticity, independence, multicollinearity | Linear Regression plots, Statistics and Save options |
| 3 Moderation | Centred predictors, interaction term, R² change, simple slopes, plot | Compute, hierarchical regression, or PROCESS Model 1 |
| 4 Discussion | Findings, implications, limitations, future research | Written section, 500 words maximum |
Part 1: Data preparation and screening
In this illustrative run, all 200 cases had complete data on the four study variables, so no case was removed. In your own report, state exactly what you found and how you handled it, with a reason. Table 1 gives the descriptives.
Table 1. Descriptive statistics (illustrative, N = 200)
| Variable | M | SD | Minimum | Maximum |
|---|---|---|---|---|
| Sleep duration (hours per night) | 6.98 | 1.18 | 4.00 | 10.00 |
| Screen time (hours per day) | 5.21 | 1.96 | 0.50 | 10.88 |
| Exercise (minutes per week) | 152.20 | 65.96 | 0.00 | 332.62 |
| Mean reaction time (seconds) | 0.430 | 0.056 | 0.266 | 0.549 |
Table 2. Pearson correlations (illustrative)
| Variable | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| 1 Sleep duration | – | |||
| 2 Screen time | –.17* | – | ||
| 3 Exercise | .09 | –.24** | – | |
| 4 Reaction time | –.34** | .33** | –.18* | – |
Note. * p < .05, ** p < .001. Reaction time is in seconds, so a negative correlation with sleep means that more sleep goes with faster responses.
Longer sleep was moderately associated with faster reaction time, r(198) = –.34, p < .001, and greater screen time with slower reaction time, r(198) = .33, p < .001. Exercise showed a small negative association with reaction time, r(198) = –.18, p = .010. In this illustration, Hypotheses 1 to 3 are therefore supported.
Part 2: Assumption testing
The assumptions of multiple regression apply to the model that contains the interaction (Field, 2024). For this illustrative model, the results were as follows.
| Assumption | Check | Illustrative result | Judgement |
|---|---|---|---|
| Normal residuals | Histogram, P-P plot, Shapiro–Wilk | W = .992, p = .293; skewness = –0.25; excess kurtosis = 0.35 | Met |
| Independent errors | Durbin–Watson | 2.07 | Met (close to 2) |
| Homoscedasticity | Residual-by-predicted plot; Breusch–Pagan | LM(3) = 0.88, p = .830 | Met |
| Multicollinearity | VIF with centred predictors | 1.04, 1.04 and 1.01 | Met |
| Outliers and influence | Standardised residuals; Cook’s distance | One case at 3.07; maximum Cook’s D = .08 | Acceptable |
Linearity was judged from the residual plot and scatterplots. A formal Breusch–Pagan test needs an extra procedure or macro in SPSS, so many reports rely on the residual plot alone. The test is listed here for completeness.
Centring matters for multicollinearity. In this illustration, using uncentred sleep, screen time and their product would have given VIF values of about 8.8, 41.0 and 42.9, whereas centring reduced them to roughly 1. This is the evidence you can cite to justify centring (Aiken & West, 1991). One case sat just beyond ±3 standardised residuals but had little influence, so it was retained. Say the same kind of thing in your report: what you found, what it implies and what you did.
If you are unsure how to read your own diagnostics, our SPSS homework help page explains how we support output checking.
Part 3: Moderation analysis
Model and SPSS syntax
Sleep duration was the predictor (X), screen time the moderator (W) and reaction time the outcome (Y). Both predictors were mean-centred before the product term was created, and the interaction was entered in a second step. The same model can be run with PROCESS Model 1 (Hayes, 2022). Use the exact means from your own Descriptives output.
COMPUTE sleep_c = sleep - 6.9847.
COMPUTE screen_c = screen - 5.2053.
COMPUTE sleep_x_screen = sleep_c * screen_c.
EXECUTE.
REGRESSION
/STATISTICS COEFF OUTS CI(95) R ANOVA CHANGE
/DEPENDENT rt_mean
/METHOD=ENTER sleep_c screen_c
/METHOD=ENTER sleep_x_screen
/SCATTERPLOT=(*ZRESID,*ZPRED)
/RESIDUALS DURBIN HISTOGRAM(ZRESID) NORMPROB(ZRESID)
/SAVE ZRESID COOK.
Table 3. Hierarchical regression predicting reaction time (illustrative)
| Predictor | b | SE | 95% CI | t(196) | p |
|---|---|---|---|---|---|
| Intercept | 0.4315 | 0.0036 | [0.4244, 0.4386] | 120.05 | < .001 |
| Sleep duration (centred) | –0.0136 | 0.0031 | [–0.0196, –0.0076] | –4.44 | < .001 |
| Screen time (centred) | 0.0083 | 0.0018 | [0.0047, 0.0120] | 4.53 | < .001 |
| Sleep × Screen time | 0.0039 | 0.0017 | [0.0007, 0.0072] | 2.37 | .019 |
Note. Step 1: R² = .194, F(2, 197) = 23.74, p < .001. Step 2: R² = .217, adjusted R² = .205, F(3, 196) = 18.07, p < .001; ΔR² = .022, F(1, 196) = 5.60, p = .019.
Cross-check the consistency: the F change for the interaction step equals the square of its t value (2.37² ≈ 5.60), and the ΔR² is the difference between the two R² values (.217 – .194).
APA 7 results write-up
A hierarchical regression tested whether screen time moderated the relationship between sleep duration and reaction time. Step 1, with centred sleep duration and screen time, explained 19.4% of the variance, F(2, 197) = 23.74, p < .001. Adding the interaction term in Step 2 produced a small but significant increase, ΔR² = .022, F(1, 196) = 5.60, p = .019, giving a final model R² = .217, F(3, 196) = 18.07, p < .001. The interaction effect corresponds to Cohen’s f² = .03, a small effect (Cohen, 1992).
Sleep duration predicted faster reaction time, b = –0.0136, 95% CI [–0.0196, –0.0076], p < .001, and screen time predicted slower reaction time, b = 0.0083, 95% CI [0.0047, 0.0120], p < .001. The interaction was positive, b = 0.0039, 95% CI [0.0007, 0.0072], p = .019. A positive coefficient means the negative link between sleep and reaction time became weaker as screen time rose.
Simple slopes and plot
Simple slopes were estimated for the effect of sleep at one standard deviation below the mean of screen time (3.25 hours per day), at the mean (5.21 hours) and one standard deviation above it (7.16 hours).
Table 4. Simple slopes of sleep duration on reaction time (illustrative)
| Screen time level | Slope (b) | SE | 95% CI | t(196) | p |
|---|---|---|---|---|---|
| Low (–1 SD; 3.25 h) | –0.0213 | 0.0043 | [–0.0297, –0.0128] | –4.98 | < .001 |
| Mean (5.21 h) | –0.0136 | 0.0031 | [–0.0196, –0.0076] | –4.44 | < .001 |
| High (+1 SD; 7.16 h) | –0.0059 | 0.0046 | [–0.0151, 0.0033] | –1.27 | .206 |
Each slope equals the sleep coefficient plus the interaction coefficient multiplied by the centred screen-time value: –0.0136 + 0.0039 × (–1.96) ≈ –0.0213 and –0.0136 + 0.0039 × 1.96 ≈ –0.0059. For your plot (Figure 1), draw reaction time on the y-axis, sleep on the x-axis and three lines for the screen-time levels. These predicted values, taken at sleep durations one standard deviation below and above the mean (5.81 and 8.16 hours), are what such a plot would show.
| Screen time level | Predicted RT at 5.81 h sleep (s) | Predicted RT at 8.16 h sleep (s) | Difference (s) |
|---|---|---|---|
| Low | 0.440 | 0.390 | 0.050 |
| Mean | 0.448 | 0.416 | 0.032 |
| High | 0.455 | 0.441 | 0.014 |
The lines converge as screen time rises. Sleep was significantly related to reaction time at low and average screen time, but not at high screen time, so Hypothesis 4 is supported in this illustration.
Want a statistician to check your own syntax and output? Our statistics assignment help in Malaysia page explains what is covered.
Part 4: Discussion and conclusion (extract)
In this illustrative example, longer sleep, lower screen time and more exercise were each linked with faster reaction time. Screen time moderated the sleep effect: the benefit of longer sleep was largest for light phone users and not statistically detectable for heavy users. A practical implication is that sleep advice may be most useful to people with moderate or low screen exposure, while heavy users may need to address screen habits as well.
Limitations belong here too. The design is cross-sectional and correlational, so causal claims are not justified. Sleep, screen time and exercise were self-reported, and reaction time was measured at home, where device and environment can add noise. The interaction effect was small (f² = .03), and the high-screen-time slope was imprecise, with a confidence interval that included zero. Future research could use device-recorded screen time, objective sleep tracking and experimental manipulation.
Students needing a tailored report can submit their assignment brief for a quote, or browse more statistics assignment samples.
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Sage Publications.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155–159. https://doi.org/10.1037/0033-2909.112.1.155
Field, A. (2024). Discovering statistics using IBM SPSS Statistics (6th ed.). SAGE Publications.
Hayes, A. F. (2022). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (3rd ed.). Guilford Press.
Prepared by
Malaysia Assignment Help Editorial Team
This model answer was prepared by our editorial team in response to a real assignment brief we received. It shows one way to approach the task and is meant as a study guide, not for submission.
