Statistical Test Selection for Pediatric Obesity Research with Multiple Diagnostic Criteria
Direct Answer
For analyzing the relationship between pediatric groups and obesity when using more than 3 diagnostic criteria (multiple outcome categories), multinomial logistic regression (Option D) is the appropriate statistical test.
Rationale for Test Selection
Why Multinomial Logistic Regression
Multinomial logistic regression is specifically designed to predict categorical outcomes with more than two categories, which directly matches your research scenario where obesity is diagnosed using multiple criteria creating several distinct weight categories 1.
The evidence consistently demonstrates that pediatric obesity research uses multiple BMI-based categories: normal weight, overweight, obesity, and severe obesity (BMI ≥99th percentile or ≥120% of 95th percentile) 1.
Studies examining pediatric populations routinely classify children into 3-4 distinct weight categories based on BMI percentiles (e.g., <5th percentile, 5th-85th percentile, 85th-95th percentile, ≥95th percentile) 1, 2.
Research specifically analyzing relationships between demographic factors and childhood obesity categories successfully employed multinomial logistic regression when the outcome variable had multiple categories 3, 4, 5.
Why Other Options Are Inappropriate
T-test (Option A) is eliminated because it only compares means between two groups and cannot handle multiple outcome categories 3.
Linear regression (Option B) is inappropriate because your outcome (obesity diagnosis) is categorical, not continuous. Linear regression requires a continuous dependent variable, whereas obesity classification creates distinct categories 3, 4.
Logistic regression (Option C) is insufficient because standard binary logistic regression only handles two outcome categories (e.g., obese vs. non-obese), but your study involves more than 3 diagnostic criteria creating multiple categories 4, 5.
Evidence from Pediatric Obesity Research
Successful Application of Multinomial Models
A study of 306 American Indian and white children (aged 8-9 years) used multinomial logistic regression to predict weight categories based on BMI percentiles, with statistically significant models for girls (χ² 20 = 42.73, P < .01), boys (χ² 20 = 50.44, P < .001), American Indian (χ² 20 = 36.67, P < .05), and white children (χ² 20 = 55.99, P < .001) 3.
Research examining 7,814 kindergarten students employed multinomial logistic regression with BMI as the dependent variable (categorized into multiple weight classes) and demographic traits, dietary practices, and physical activity as independent variables 4.
A cross-sectional study of 1,317 children aged 2-16 years used multiple logistic regression to study the relationship between obesity/overweight categories and different variables, calculating adjusted odds ratios for each weight category 5.
Clinical Context Supporting Multiple Categories
Standard Pediatric Weight Classifications
The American Heart Association guidelines establish that pediatric obesity assessment requires multiple BMI percentile categories: underweight (<5th percentile), normal weight (5th-85th percentile), overweight (85th-95th percentile), and obesity (≥95th percentile) 1, 2.
Severe obesity represents an additional category defined as BMI ≥99th percentile or ≥120% of the 95th percentile, creating at least 4-5 distinct diagnostic categories in comprehensive pediatric obesity research 1.
Studies consistently demonstrate different risk factor profiles across these multiple weight categories, with severe obesity showing 50% prevalence of metabolic syndrome clustering compared to 0% in normal-weight participants 1.
Multiple Objectives Require Categorical Analysis
When research examines relationships between pediatric groups (age, sex, race/ethnicity) and obesity with multiple diagnostic criteria, the outcome naturally becomes polytomous (multiple unordered categories) 1.
Prevalence data show distinct patterns across weight categories: among 2-19 year-olds, 3.1% of white children, 5.2% of Hispanic children, and 5.8% of black children had BMI ≥99th percentile, demonstrating the need to analyze multiple outcome categories simultaneously 1.
Common Pitfalls to Avoid
Do not collapse multiple obesity categories into binary outcomes (obese vs. non-obese), as this loses critical information about the gradient of risk across weight categories and reduces statistical power 1.
Avoid using multiple binary logistic regressions for each category comparison, as this inflates Type I error and doesn't account for the relationships between all categories simultaneously 3, 4.
Ensure adequate sample size in each outcome category, as multinomial logistic regression requires sufficient observations across all categories for stable estimates 3, 5.