Statistical Test Selection for Pediatric Obesity Research with Multiple Diagnostic Criteria
For analyzing the relationship between pediatric groups and obesity when multiple diagnostic criteria (more than 3 categories) are used, multinomial logistic regression (Option D) is the correct statistical test. 1
Why Multinomial Logistic Regression is Required
Multinomial logistic regression is specifically designed to handle categorical outcomes with more than two categories, making it essential when obesity is classified into multiple groups based on different diagnostic criteria. 1 This approach allows simultaneous analysis of how pediatric groups relate to each obesity category while maintaining the distinct clinical classifications that are critical for pediatric obesity assessment. 1
Key Advantages for This Research Design
Handles multiple outcome categories simultaneously without collapsing clinically meaningful distinctions between obesity severity levels (e.g., normal weight, overweight, Class I obesity, Class II obesity, Class III obesity). 1
Preserves the categorical nature of obesity classification based on BMI percentiles, which creates distinct groups rather than a continuous variable in pediatric populations. 1
Allows examination of multiple independent variables (pediatric group characteristics) while accounting for their unique contributions to each obesity category. 2
Why Other Options Are Inappropriate
T-test (Option A) - Incorrect
- T-tests only compare means between two groups and cannot handle multiple categorical outcomes or multiple predictor variables simultaneously. 1
- This test is designed for continuous dependent variables with binary grouping, not for examining relationships with multiple obesity categories. 1
Linear Regression (Option B) - Incorrect
- Linear regression requires a continuous dependent variable, but obesity classification based on BMI percentiles creates categorical groups (normal, overweight, obese Class I/II/III), not a continuous outcome. 1
- Using linear regression would violate fundamental assumptions about the nature of the outcome variable. 1
Logistic Regression (Option C) - Insufficient
- Standard logistic regression only handles binary outcomes (two categories), requiring researchers to collapse multiple obesity categories into just two groups. 1
- This approach loses important clinical distinctions between different severity levels of obesity that are essential for pediatric assessment and treatment planning. 1
- The Edmonton Obesity Staging System for Pediatrics (EOSS-P) and other staging systems specifically identify multiple categories (Stages 0-4) that should not be collapsed. 3
Clinical Context Supporting This Choice
Pediatric obesity assessment inherently involves multiple categories: BMI percentiles create distinct groups (≥85th percentile for overweight, ≥95th percentile for obesity), and staging systems like EOSS-P further stratify children into multiple risk categories. 3
Meta-analyses of pediatric obesity use appropriate statistical methods for categorical outcomes, such as pooled risk differences for obesity prevalence changes, emphasizing the importance of selecting methods that match the outcome structure. 3
Multiple diagnostic criteria in pediatric obesity research may include BMI percentile categories, waist circumference classifications, metabolic complication staging, and functional limitation categories—all creating multiple outcome groups that require multinomial analysis. 3
Common Pitfall to Avoid
Do not collapse multiple obesity categories into binary outcomes (obese vs. non-obese) to use standard logistic regression, as this eliminates clinically important distinctions between overweight, mild obesity, and severe obesity that guide different treatment intensities and predict different health outcomes. 3, 1