Research Methods and Statistics is the smallest of the eight Part 1 domains, 7% of the Examination for Professional Practice in Psychology (EPPP) knowledge exam, but its questions reward exact reasoning. Six of these 10 are about choosing or reading an analysis: a MANOVA for three correlated outcomes, a mixed factorial ANOVA for two groups measured four times, hierarchical regression for a theory-driven order of entry, multicollinearity at r = .92, a forest-plot diamond that crosses zero and trim-and-fill after a significant Egger’s test. The other four are a Type II error, lowering alpha from .05 to .01 with 40 participants, a selection-maturation threat in two intact wards and path analysis with 45 skewed cases.
Question 1
A psychologist is building a predictive model for depression severity using several baseline clinical measures. During the analysis, the researcher notices that two predictors correlate at r = .92 and that the beta weights fluctuate wildly when tested on different subsamples. Based on these findings, what is the most likely statistical consequence for this regression model?
- A) The overall model R-squared will artificially decrease.
- B) The standard errors of the coefficients will inflate.✓
- C) The criterion variable will violate the normality assumption.
- D) The homoscedasticity of the residual variance will improve.
Show rationale
When predictor variables are highly correlated, it creates a condition known as multicollinearity. This makes it difficult for the model to isolate the unique contribution of each variable, which directly inflates the standard errors of the regression coefficients and leads to unstable beta weights across samples. Option B is correct because this is the hallmark mathematical consequence of multicollinearity. Option A is incorrect; multicollinearity does not typically decrease the overall R-squared, as the shared variance still contributes to the model's total predictive power. Option C is incorrect because multicollinearity involves relationships among predictors, not the normality of the criterion variable itself. Option D is incorrect because homoscedasticity relates to the uniform spread of residuals, which is entirely unrelated to high correlations between independent variables. Key Takeaway: Multicollinearity inflates standard errors and destabilizes beta weights without necessarily reducing overall predictive power.
Question 2
A neuropsychological researcher investigates whether a novel cognitive training program improves working memory in older adults. Unknown to the investigator, the intervention genuinely produces a meaningful cognitive benefit, but the statistical analysis yields a non-significant result. What research error has occurred?
- A) A Type II error resulting from mistakenly retaining a false null hypothesis.✓
- B) A Hawthorne effect resulting from participants altering behavior due to being observed.
- C) A Type I error resulting from mistakenly rejecting a true null hypothesis.
- D) A Berkson bias resulting from non-random hospital selection of clinical research participants.
💡 Key Takeaway
A Type II error occurs when a statistical test fails to reject a false null hypothesis, missing a real treatment effect.
Show rationale
In inferential statistical hypothesis testing, a Type II error (beta error) occurs when a researcher fails to reject (i.e., retains) a null hypothesis that is actually false. In this scenario, the cognitive training intervention truly produces a meaningful clinical benefit, meaning a true effect exists in the population and the null hypothesis is false. Because the statistical test yielded a non-significant result, the researcher concluded there was no significant effect, erroneously retaining the false null hypothesis and committing a Type II error. In contrast, a Type I error occurs when a researcher mistakenly rejects a null hypothesis that is actually true, falsely claiming an intervention effect or association when none exists in the population. Berkson bias (or Berkson's fallacy) is incorrect because it is a specific form of selection bias arising when hospitalized or clinical patients are sampled, leading to spuriously distorted associations between diseases or conditions. Finally, the Hawthorne effect is incorrect because it describes an observational reactivity artifact wherein study participants unconsciously alter their behavioral or cognitive performance simply because they are aware of being monitored.
Question 3
A researcher compares a cognitive-behavioral intervention implemented in one intact psychiatric unit with standard care in another unit. Despite similar baseline symptom severity, the researcher notes a lack of random assignment between the two hospital wards. Which threat to internal validity is most prominent?
- A) Statistical regression to the mean
- B) Selection-maturation interaction effects✓
- C) Reactive effects of the pretesting
- D) Experimental mortality and attrition bias
Show rationale
When using intact groups without random assignment, a selection-maturation interaction is a major threat because the groups may naturally develop or recover at different rates, even if baseline scores appear similar. Statistical regression applies primarily to extreme scores, not necessarily intact groups. Reactive pretesting and mortality bias might occur, but the fundamental flaw of nonequivalent control groups is the interaction between selection and other developmental threats. Key Takeaway: Selection-maturation interaction is a primary threat to internal validity in nonequivalent control group designs.
Question 4
A clinical researcher wants to control for age and baseline severity first before attempting to determine if a novel biomarker adds variance to treatment response. Because there is a strict theory-driven order of variable entry, which statistical approach is most appropriate for this analysis?
- A) Conduct a forward stepwise multiple regression analysis.
- B) Perform a standard simultaneous multiple regression analysis.
- C) Execute a hierarchical sequential multiple regression analysis.✓
- D) Utilize a backward elimination multiple regression analysis.
Show rationale
When a researcher needs to control for specific covariates based on a theoretical rationale before evaluating new predictors, hierarchical regression is the correct method. In this approach, the researcher manually dictates the blocks of entry, allowing them to see the exact change in R-squared when the novel biomarker is added after age and baseline severity. Option C is therefore correct. Options A and D are incorrect because stepwise regression methods (both forward and backward) rely entirely on mathematical algorithms and statistical significance to determine the order of entry, stripping the researcher of theoretical control. Option B is incorrect because simultaneous regression enters all variables into the model at the exact same time, making it impossible to evaluate the incremental predictive value of the biomarker after controlling for the baseline covariates. Key Takeaway: Hierarchical regression is used when the researcher has a theoretical rationale for the specific order of predictor entry.
Question 5
A meta-analysis examining telepsychology efficacy yields a significant Egger's regression test. The researchers want to mathematically estimate what the effect size would be if the missing unpublished studies were actually included in the analysis. Which procedure should they utilize?
- A) Conduct a trim-and-fill procedure to adjust the pooled estimate.✓
- B) Calculate Cochran’s Q statistic to adjust the pooled estimate.
- C) Generate a standard forest plot to visualize the missing data.
- D) Perform a meta-regression analysis to visualize the missing data.
Show rationale
The correct answer is A because the trim-and-fill procedure is specifically designed to impute missing studies to correct for funnel plot asymmetry, allowing researchers to estimate an adjusted pooled effect size that accounts for publication bias. Option B is incorrect because Cochran’s Q evaluates heterogeneity, not missing data or publication bias. Option C is incorrect because while a standard forest plot visualizes the effect sizes of known studies, it cannot mathematically estimate or visualize missing data. Option D is incorrect because meta-regression is used to explore how specific moderator variables influence the effect size across studies, rather than adjusting for publication bias or imputing missing studies. Key Takeaway: The trim-and-fill procedure imputes missing studies to adjust pooled effect sizes for publication bias.
Question 6
A psychologist is designing a clinical trial to evaluate a novel exposure therapy for PTSD. To ensure they do not endorse an ineffective treatment, the researcher changes the alpha level from .05 to .01 while maintaining a fixed sample size of 40 participants. What is the most likely statistical consequence of this design modification?
- A) It decreases Type I error risk but increases the likelihood of a Type II error.✓
- B) It decreases Type II error risk but increases the likelihood of a Type I error.
- C) It increases Type I error risk and increases the likelihood of a Type II error.
- D) It increases Type II error risk and decreases the likelihood of a Type I error.
Show rationale
By lowering the alpha level from .05 to .01, the researcher makes the criteria for statistical significance much stricter. This directly reduces the risk of a Type I error (false positive), meaning they are less likely to claim the therapy works when it actually does not. However, because the sample size remains fixed, this stricter threshold simultaneously reduces statistical power. As power decreases, the risk of a Type II error (false negative) inherently increases, meaning the researcher is now more likely to miss a true treatment effect. Option A correctly identifies this inverse relationship. Options B, C, and D misrepresent this fundamental trade-off; you cannot decrease the risk of both errors simultaneously without increasing the sample size, and lowering alpha never increases Type I error risk. Key Takeaway: Decreasing the alpha level reduces false positives but inherently increases the risk of false negatives.
Question 7
A clinical researcher is evaluating a new trauma intervention by comparing three distinct treatment groups (CBT, EMDR, and Waitlist). To capture the full clinical picture, the researcher measures outcomes using three highly correlated outcome measures assessing PTSD, anxiety, and depression. Which statistical approach is most appropriate for analyzing this data?
- A) Run one MANOVA to strictly control the overall family-wise error rate.✓
- B) Run three ANOVAs to strictly control the overall family-wise error rate.
- C) Run one factorial ANOVA to evaluate the potential treatment interaction effects.
- D) Run three independent t-tests to evaluate the potential treatment interaction effects.
Show rationale
The researcher is evaluating three highly correlated outcome measures across distinct groups, which necessitates a multivariate approach. A MANOVA is the most appropriate statistical test because it analyzes multiple continuous dependent variables simultaneously while controlling the family-wise error rate. If the researcher were to run three separate ANOVAs (Option B), the risk of committing a Type I error would inflate significantly due to multiple testing. Option C is incorrect because a factorial ANOVA is used for multiple independent variables, not multiple dependent variables. Option D is incorrect because independent t-tests only compare two groups on a single variable and would further inflate error rates. By using a single MANOVA, the researcher ensures statistical rigor when analyzing complex, correlated clinical outcomes. Key Takeaway: MANOVA is used to analyze multiple correlated dependent variables simultaneously while controlling family-wise error.
Question 8
A researcher evaluates a forest plot from a meta-analysis comparing eye movement desensitization and reprocessing to trauma-focused CBT. The pooled effect size diamond at the bottom of the plot crosses the vertical line at zero. How should the researcher interpret this specific visual finding?
- A) The combined studies demonstrate a significant difference between treatments.
- B) The individual studies indicate severe publication bias within the literature.
- C) The combined studies fail to show a significant treatment difference.✓
- D) The individual studies possess high methodological heterogeneity across samples.
Show rationale
The correct answer is C because in a forest plot displaying mean differences, the vertical line at zero represents the line of no effect. When the confidence interval of the pooled effect size (the diamond) crosses this line, the overall result is not statistically significant. Option A is incorrect because crossing zero explicitly indicates a lack of significant difference. Option B is incorrect as publication bias is typically evaluated using a funnel plot, not the pooled diamond on a forest plot. Option D is incorrect because while individual study lines might indicate heterogeneity if they vary widely, the diamond crossing zero specifically denotes the lack of a significant combined treatment effect, not methodological variance. Key Takeaway: In a forest plot of mean differences, a pooled diamond crossing zero indicates non-significant results.
Question 9
A clinical psychologist attempts to conduct a path analysis evaluating treatment outcomes using a sample size of forty-five patients and highly skewed self-report symptom data. Based on the fundamental assumptions of structural modeling techniques, what is the primary statistical risk of proceeding with this specific dataset?
- A) Generating inflated reliability estimates for the scales
- B) Producing biased parameter estimates and fit indices✓
- C) Creating excessive multicollinearity among the predictor variables
- D) Forcing endogenous variables to become exogenous factors
Show rationale
The primary risk is producing biased parameter estimates and fit indices because path analysis relies on maximum likelihood estimation, which assumes multivariate normality and requires a large sample size to function accurately. Generating inflated reliability estimates for the scales is incorrect because sample size and non-normality primarily distort standard errors and model fit rather than artificially inflating scale reliability. Creating excessive multicollinearity among the predictor variables is incorrect because multicollinearity is a property of the relationships between variables, not a byproduct of small or skewed samples. Forcing endogenous variables to become exogenous factors is incorrect because variable specification is determined by the theoretical model drawn by the researcher, not altered by sample characteristics. Key Takeaway: Path analysis and SEM require large sample sizes and multivariate normality to produce unbiased estimates and reliable fit indices.
Question 10
A clinical research team wants to evaluate the effectiveness of two distinct interventions for insomnia. They measure sleep quality in two distinct treatment groups (Sleep Restriction Therapy vs. Sleep Hygiene Education) assessed repeatedly at four distinct time points over twelve months. Which statistical analysis is most appropriate for this study design?
- A) Run a one-way independent ANOVA to analyze the overall group differences.
- B) Run a one-way repeated measures ANOVA to analyze the group differences.
- C) Run a simple multivariate analysis of variance to analyze group differences.
- D) Run a mixed factorial ANOVA to analyze the interaction and differences.✓
Show rationale
The study design involves two distinct treatment groups being measured repeatedly at four distinct time points over twelve months. This requires a mixed factorial ANOVA, which simultaneously evaluates a between-subjects factor (treatment group) and a within-subjects factor (time). Option A is incorrect because a one-way independent ANOVA only analyzes between-subjects differences at a single time point, ignoring the repeated measures. Option B is incorrect because a one-way repeated measures ANOVA only analyzes changes over time for a single group, ignoring the distinct treatment groups. Option C is incorrect because a simple MANOVA is used for multiple dependent variables at one time point, rather than a single dependent variable measured repeatedly across mixed groups. Key Takeaway: A mixed factorial ANOVA analyzes data containing both between-subjects and within-subjects (repeated measures) factors.
EPPP Part 1 Simulator Pro
What the full simulator adds
- ✓Each explanation works through why the keyed answer holds and where the other choices go wrong, so the concept sticks for the next item
- ✓None of the 1,923 is an ASPPB exam item; each one is filed under one of the eight Part 1 domains ASPPB uses for exams given in 2026 and 2027
- ✓Ten tests of 175 questions and one of 173, timed at 198 minutes: the scored-item count at the real 225-item, 4 h 15 min pace. Ours leave out the 50 unscored items and always offer four choices, where real items have three or four
- ✓A score for each of the eight domains, plus an overall percentage measured against a 70% readiness line we set ourselves; ASPPB reports a scaled 200–800 score and recommends 500 to pass
$29.99 lifetime · or $9.99/month
🛡️ 30-day money-back guarantee · secure PayPal checkout · cancel anytime
📚 10+ years publishing exam-prep books · 150,000+ books sold✅ ASPPB EPPP Part 1 exam outline · current for 2026