- Which statistical test is most appropriate to determine if there is a significant difference in the mean IOP between the medication and placebo groups?
Answer: Independent samples t-test
An independent samples t-test (or two-sample t-test) is the correct statistical method for comparing the means of a continuous outcome variable (IOP) between two independent, unrelated groups (medication vs. placebo). A paired t-test is used for related samples (e.g., before and after treatment in the same group). ANOVA is used for comparing means of three or more groups. Pearson correlation assesses the association between two continuous variables.
- A cardiology database captures New York Heart Association (NYHA) heart-failure class I–IV at each annual visit for thousands of patients. Investigators want to test whether there is a monotonic trend toward higher NYHA class with increasing quartiles of serum brain natriuretic peptide (BNP) concentration. Which statistical approach is most appropriate?
Answer: Spearman rank correlation testing ordinal trend between BNP quartile and NYHA class
When you encounter questions about analyzing the relationship between two ordinal variables, you need to identify both the data types and the specific research question being asked.
In this scenario, you have two ordinal variables: NYHA class (I-IV, representing increasing severity) and BNP quartiles (1st-4th, representing increasing concentration). The investigators specifically want to test for a "monotonic trend" - whether higher BNP quartiles consistently correspond to higher NYHA classes.
Answer A is correct because Spearman rank correlation is designed exactly for this situation. It tests for monotonic relationships between ordinal variables by ranking the data and calculating correlation coefficients. Since both variables are ordinal and the question asks about directional trend, this is the ideal statistical approach.
Answer B is wrong because chi-square tests of independence treat variables as nominal (categorical without order), ignoring the inherent ranking in both NYHA class and BNP quartiles. This wastes valuable information about the ordered nature of your data.
Answer C is incorrect because paired t-tests compare means of continuous variables within the same subjects over time. Here, you're looking at the relationship between two different ordinal measures, not comparing means of the same variable.
Answer D is wrong because log-rank tests compare survival curves between groups over time. This study isn't analyzing time-to-event data or survival outcomes.
Study tip: When you see ordinal data and questions about "trends" or "associations," think Spearman correlation. When you see nominal categories, think chi-square. Always match your statistical test to your data types and research question.
- The use of an independent samples t-test in this situation may be invalid because which of its key assumptions has been violated?
Answer: The data from the underlying population must be normally distributed.
The t-test is a parametric test, which means it relies on certain assumptions about the data. A key assumption is that the data are sampled from a population with a normal distribution. While the t-test is robust to minor violations, especially with larger sample sizes, significant skewness in a small sample (n=12) makes its results unreliable. The assumption of independence is met, and the outcome is continuous. While equal variances (homoscedasticity) is an assumption, the violation of normality is the more critical issue here prompting the use of a non-parametric test.
- A nutrition researcher compares mean fasting LDL cholesterol across four independent diet groups (Mediterranean, DASH, Ketogenic, and Control). LDL values are continuous and satisfy normality and equal variance assumptions. Which statistical test should be used to evaluate whether at least one diet differs from the others?
Answer: One-way analysis of variance for four independent means
When you encounter a question comparing means across multiple independent groups with continuous data, you're in ANOVA territory. The key decision factors are: number of groups (more than two), data type (continuous), independence of groups, and whether parametric assumptions are met.
Why A is correct: One-way ANOVA is specifically designed to compare means across multiple independent groups when you have continuous data that meets parametric assumptions (normality and equal variance). The "one-way" refers to having one independent variable (diet type) with multiple levels (four diets). ANOVA tests the null hypothesis that all group means are equal against the alternative that at least one mean differs from the others—exactly what this researcher wants to know.
Why the other options are wrong: Choice B (Chi-square) is used for categorical data and frequency comparisons, not continuous measurements like LDL cholesterol levels. Choice C (Kruskal-Wallis) is the non-parametric alternative to one-way ANOVA, used when normality or equal variance assumptions are violated—but the question explicitly states these assumptions are satisfied. Choice D (paired t-test) is for comparing two related groups or repeated measurements on the same subjects, not four independent diet groups with different participants.
USMLE Strategy: When you see multiple independent groups (>2) with continuous data meeting parametric assumptions, think one-way ANOVA. If assumptions were violated, you'd use Kruskal-Wallis. For just two groups, you'd use an independent t-test. The exam often tests whether you can distinguish between parametric and non-parametric tests based on assumption violations.
- Which statistical method is most appropriate for building this predictive model?
Answer: Logistic regression
Logistic regression is the appropriate statistical technique when the dependent (outcome) variable is dichotomous or binary, such as the occurrence of an MI (yes/no). It allows for the modeling of how multiple independent variables (both continuous and categorical) predict the probability of the outcome. Multiple linear regression is used for continuous outcomes. A Cox model would be used for time-to-event data, which is more specific than what is described.
- Which of the following is the best interpretation of this correlation coefficient?
Answer: Higher BMI is associated with higher fasting glucose levels in this group.
The Pearson correlation coefficient (r) measures the strength and direction of a linear relationship between two continuous variables. A positive value (r = +0.60) indicates a positive association, meaning that as one variable (BMI) increases, the other variable (fasting glucose) tends to increase. The slope of the relationship is determined by regression analysis, not the correlation coefficient itself. The proportion of variance explained is r-squared (r²), which would be 0.36 or 36% in this case. Correlation does not establish causation.
- What is the most appropriate statistical test to use to compare the central tendency of time to pain relief among the three groups?
Answer: Kruskal-Wallis test
The Kruskal-Wallis test is the nonparametric alternative to one-way ANOVA. It is used to determine if there are statistically significant differences between two or more independent groups on a continuous or ordinal dependent variable when the assumptions of ANOVA (like normality) are not met. Since the data are highly skewed and there are three independent groups, the Kruskal-Wallis test is the most suitable choice.
- A neurophysiology study records nerve conduction velocity (m/s) and patient serum potassium concentration (mmol/L). Both variables are continuous and approximately normally distributed. Which test is most appropriate to quantify the strength and direction of the linear relationship between these two variables?
Answer: Pearson product–moment correlation coefficient
When you encounter questions about analyzing relationships between two continuous variables, you need to identify the appropriate statistical test based on the data characteristics and research question.
The Pearson product-moment correlation coefficient (choice A) is the gold standard for quantifying linear relationships between two continuous, normally distributed variables. Since both nerve conduction velocity and serum potassium are continuous and normally distributed, Pearson correlation directly measures the strength (-1 to +1) and direction (positive or negative) of their linear association.
Choice B, Spearman rank correlation, is used when data violates normality assumptions or when measuring monotonic (but not necessarily linear) relationships. Since your data is normally distributed, this non-parametric alternative isn't necessary.
Choice C, simple linear regression testing if slope equals zero, actually uses the same underlying mathematics as Pearson correlation but answers a different question. While regression can test for a relationship, it's primarily designed for prediction rather than simply quantifying association strength and direction.
Choice D, chi-square goodness-of-fit, compares observed versus expected frequencies in categorical data. This test requires binning continuous data into categories, which loses information and isn't appropriate for examining linear relationships between continuous variables.
Study tip: Remember the "two continuous, normal" pattern - when you see these characteristics with a question about relationship strength and direction, think Pearson correlation first. Save Spearman for non-normal data or ordinal variables, and reserve regression for prediction questions.
- Paired binary outcome before/after intervention. Which test?
Answer: McNemar's test
For a before/after binary outcome, the two observations per subject are paired, so you must use McNemar's test to compare discordant pairs. The chi-square test is tempting because both variables are categorical, but it treats the groups as independent and ignores the pairing. Paired t-test is wrong because the outcome is binary, not continuous.
- An endocrinology investigator records fasting glucose levels in 40 patients at baseline, 1 month, and 3 months after initiation of a dietary intervention. All measurements are normally distributed. Which statistical method best assesses whether mean glucose differs across the three time points within the same subjects?
Answer: Repeated-measures one-way ANOVA for correlated observations
When analyzing repeated measurements on the same subjects over time, you need to account for the correlation between observations from each individual. This scenario involves measuring glucose in the same 40 patients at three different time points, creating dependent (correlated) data.
Choice A is correct because repeated-measures one-way ANOVA is specifically designed for this situation. It compares means across multiple time points while accounting for the fact that measurements come from the same subjects. The "correlated observations" designation is key—it recognizes that a patient's glucose at month 1 is likely correlated with their glucose at baseline and month 3. This test maintains statistical power while controlling for individual patient variability.
Choice B is wrong because unpaired one-way ANOVA assumes independent groups, treating each measurement as if it came from different subjects. This violates the assumption of independence and would ignore the valuable information about individual patient changes over time.
Choice C is wrong because the Wilcoxon signed-rank test only compares two paired time points, but you have three time points to analyze simultaneously. Additionally, since the data is normally distributed, a parametric test (ANOVA) is more appropriate and powerful than this non-parametric alternative.
Choice D is wrong because chi-square tests analyze categorical data relationships, not continuous glucose measurements. Converting continuous glucose values into categories would also result in loss of statistical information and power.
Study tip: When you see "same subjects measured multiple times" with normally distributed data, think repeated-measures ANOVA. The key phrase "within the same subjects" signals correlated data requiring specialized analysis.
- An investigator wishes to evaluate whether the incidence rate of hospital-acquired Clostridioides difficile infection changed after a new cleaning protocol was instituted. Hospital census data provide total patient-days for 12 months before and 12 months after the intervention. Which statistical test best compares the incidence rates (cases per 1,000 patient-days) between the two periods?
Answer: Incidence rate ratio using mid-P exact or Poisson test for counts over exposure time
When comparing infection rates over time periods with different exposure durations, you're dealing with count data that follows a Poisson distribution. The key insight is that you have cases occurring over patient-days of exposure, making this a classic incidence rate comparison problem.
Answer A is correct because incidence rates represent counts per unit time (or exposure), which naturally follow a Poisson distribution. The incidence rate ratio compares rates between two periods while properly accounting for the underlying statistical distribution of rare events like healthcare-associated infections. The mid-P exact test or Poisson test specifically handles count data over exposure time, making it the gold standard for this type of epidemiological comparison.
Answer B is wrong because the chi-square test treats each month categorically and doesn't account for varying exposure times (patient-days) or the count nature of infection data. It would ignore the crucial denominator information.
Answer C is incorrect because a Student t-test assumes normally distributed continuous data and doesn't adjust for exposure time. Simply comparing mean monthly case counts ignores that months might have different patient-days, making raw counts misleading.
Answer D is flawed because the Wilcoxon signed-rank test is for paired, non-parametric comparisons of ranks, not count data. While pairing by season might control for seasonal variation, this test doesn't properly handle incidence rates or count distributions.
Remember: When you see "incidence rate" or "cases per exposure time" in epidemiology questions, think Poisson distribution and rate ratios. The exposure denominator is crucial and rules out tests that only compare raw counts.
- Pain intensity is scored on an ordinal 0–10 visual analog scale in patients receiving one of three opioid regimens. The distributions are markedly non-normal despite similar medians. Which test is most appropriate to compare the pain score distributions across the three independent groups?
Answer: Kruskal–Wallis test, the non-parametric analogue of one-way ANOVA
When you encounter a question about comparing groups with non-normal distributions and ordinal data, you need to think about parametric versus non-parametric statistical tests. The key clues here are the non-normal distributions and the ordinal nature of the 0-10 pain scale.
The Kruskal-Wallis test (A) is the correct choice because it's specifically designed for comparing three or more independent groups when the data doesn't meet the assumptions for parametric tests. This non-parametric test works by ranking all observations across groups and comparing the rank distributions, making it ideal for ordinal data with non-normal distributions.
One-way ANOVA (B) is inappropriate because it requires normally distributed residuals and homogeneity of variance - assumptions that are explicitly violated in this scenario. Using ANOVA with non-normal data would produce unreliable results.
The Mann-Whitney U test (C) is designed for comparing only two groups, not three. While you could theoretically perform multiple pairwise comparisons, this approach increases Type I error risk and doesn't provide an overall test of group differences. You'd need post-hoc corrections, making it unnecessarily complicated.
The Pearson chi-square test (D) would require converting the continuous ordinal scores into categorical bins, which loses valuable information about the magnitude of differences between scores. This conversion is neither necessary nor optimal when better tests are available.
Remember this pattern: for three or more independent groups with ordinal or non-normal data, think Kruskal-Wallis. For two groups with similar conditions, think Mann-Whitney U. Always check distribution assumptions before choosing parametric tests.
- Which of the following statistical methods is most appropriate for developing an equation to estimate GFR from serum creatinine?
Answer: Simple linear regression
Simple linear regression is used to model the relationship between two continuous variables and to create a predictive equation. The goal is to predict the value of a dependent variable (GFR) based on the value of an independent variable (serum creatinine). The other options are incorrect: a t-test compares means between two groups, ANOVA compares means among three or more groups, and a chi-square test assesses the association between categorical variables.
- Which statistical test is most suitable for determining if there is a significant association between the presence of Allele A and the diagnosis of Crohn's disease?
Answer: Chi-square test
The Chi-square test is used to assess for an association between two categorical variables. In this scenario, both the genetic polymorphism (present/absent) and the disease status (diagnosed/not diagnosed) are categorical. The test compares the observed frequencies in each category to the frequencies that would be expected if there were no association between the variables.
- Compare ordinal disability scores across 3 treatments. Which test?
Answer: Kruskal-Wallis
The outcome is ordinal and there are three independent treatment groups, so use Kruskal-Wallis, the nonparametric rank-based alternative to one-way ANOVA. One-way ANOVA is tempting but wrongly assumes a continuous, normally distributed outcome rather than ordinal disability scores.
- Based on this information, which of the following is the most appropriate conclusion?
Answer: The observed association is likely due to a confounding variable.
Correlation does not imply causation. A strong correlation between two variables does not mean that one causes the other. In this classic example, a third factor, or confounding variable (e.g., warm weather), is likely responsible for the increase in both ice cream sales and swimming activities, which in turn leads to more drowning incidents. This is a much more plausible explanation than a direct causal link.
- Given the non-normal distribution of the data, which of the following is the most appropriate statistical test?
Answer: Mann-Whitney U test
The Mann-Whitney U test is the nonparametric alternative to the independent samples t-test. It is used to compare two independent groups when the dependent variable is ordinal or when it is continuous but not normally distributed. Since the anxiety scores are not normally distributed and the groups (drug vs. placebo) are independent, the Mann-Whitney U test is the most appropriate choice. The Kruskal-Wallis test is used for three or more groups.
- Quantify linear association of two continuous variables. Which test?
Answer: Pearson correlation
To quantify linear association between two continuous variables, use Pearson correlation; it measures the strength and direction of a linear relationship. Spearman correlation is the tempting choice, but it assesses monotonic association, not specifically linearity. Chi-squared and Mann-Whitney apply to categorical or group comparisons.
- A clinical microbiology lab tests whether a novel rapid assay's ability to detect methicillin-resistant Staphylococcus aureus (MRSA) differs from the standard culture method in a sample of 15 wound specimens. Because several expected cell counts are <5, which test should be used to compare the two diagnostic methods?
Answer: Fisher exact test for small 2×2 contingency tables with low expected frequencies
When you encounter statistical test selection questions, focus on three key factors: the type of data, the study design, and whether assumptions for common tests are met.
This study compares two diagnostic methods (novel assay vs. standard culture) on the same 15 specimens, creating a 2×2 contingency table with paired data. The critical detail is that "several expected cell counts are <5," which violates the assumption for chi-square tests that all expected frequencies should be ≥5.
Answer A is correct because Fisher's exact test is specifically designed for small 2×2 contingency tables where expected frequencies are too low for chi-square approximation. Unlike chi-square, Fisher's exact test calculates exact probabilities without relying on large-sample approximations, making it ideal when cell counts are <5.
Answer B is wrong because continuing with chi-square despite small expected counts would violate the test's fundamental assumption and could lead to inaccurate p-values. The question explicitly states this assumption is violated.
Answer C is tempting since McNemar's test does handle paired diagnostic comparisons, but it's used when you're specifically testing for discordant pairs (when one test is positive and the other negative). The question doesn't indicate this specific focus on disagreement patterns.
Answer D is incorrect because both diagnostic methods produce categorical results (positive/negative for MRSA), not continuous accuracy scores that would warrant a t-test.
Remember: When expected cell counts are <5 in a 2×2 table, always choose Fisher's exact test over chi-square. This is a common USMLE trap testing your knowledge of statistical test assumptions.
- Which statistical test is the most appropriate to compare the mean CRP levels between the two groups?
Answer: Independent samples t-test
This study compares the means of a continuous, normally distributed variable (CRP levels) between two distinct, unrelated groups (RA patients and healthy controls). Therefore, the independent samples t-test is the most appropriate statistical method. A paired t-test is for related samples, a chi-square test is for categorical data, and Pearson correlation assesses the relationship between two continuous variables.
