Examkin

Question

Difficulty: MediumInferential Statistical Tests

An informatics board at a tertiary care center is evaluating a newly implemented machine learning algorithm designed to detect acute appendicitis on abdominal ultrasound. To validate the algorithm's performance, a dataset of 150 historical ultrasound scans with known surgical outcomes is assembled. Each of the 150 scans is independently evaluated by both the machine learning algorithm and a senior attending radiologist. The diagnostic result for each scan is recorded categorically as either 'positive' or 'negative' for appendicitis by each evaluator. The board aims to determine if there is a statistically significant difference in the overall proportion of positive diagnoses made by the algorithm compared to the radiologist on these identical scans. Which of the following is the most appropriate statistical test to analyze these data?

  1. A
    Pearson chi-square test
  2. B
    Fisher exact test
  3. McNemar testAnswer
  4. D
    Paired t-test
  5. E
    Wilcoxon signed-rank test

Answer

McNemar test
The correct statistical test is the McNemar test. This study compares a categorical outcome ('positive' or 'negative') between two groups. The groups are paired because the same set of 150 ultrasound scans is being evaluated by two different raters (the machine learning algorithm and the human radiologist). The McNemar test is specifically designed to analyze paired nominal data to determine if there is a statistically significant difference in marginal frequencies (proportions) between two related groups.

Step-by-Step Solution

1
Identify the type of dependent variable being measured.
The outcome variable is categorical (nominal), specifically binary ('positive' or 'negative' for appendicitis).
Determining the variable type (categorical vs. continuous) is the essential first step in selecting a statistical test.
2
Determine if the study groups are independent or paired.
The groups are paired because the exact same 150 ultrasound scans are being evaluated by both diagnostic methods.
Statistical test selection changes depending on whether measurements are drawn from independent samples or are repeated measures on the same subjects/items.
3
Select the appropriate test for paired categorical data.
The McNemar test is chosen.
The McNemar test is specifically designed to compare differences in proportions for paired nominal data.

Key Concept

Selection of statistical tests for paired categorical data
Rate this question

Topics

Inferential Statistical Tests

More questions from this topic