Letter to the editor: robustness of osteoporosis risk prediction models with enhanced statistical analyses

Pires de Carvalho, Filipe Ricardo; Gavaia, Paulo

http://hdl.handle.net/10400.1/28549

Utilize este identificador para referenciar este registo.

Nome:	Descrição:	Tamanho:	Formato:
1-s2.0-S0010482525010625-main.pdf		323.53 KB	Adobe PDF	Ver/Abrir

Contacte-nos

Autores

Pires de Carvalho, Filipe Ricardo

Gavaia, Paulo

Resumo(s)

In response to Oka et al.’s letter, we conducted additional statistical analyses to validate the robustness of our osteoporosis risk prediction model using NHANES 2007–2014 data (n = 7924). We evaluated 10 key predictors through Spearman’s rho, Kendall’s tau, Mutual Information (MI), and Total Correlation. Weight (BMX_BMXWT) and arm circumference (BMX_BMXARMC) showed strong negative correlations with osteoporosis risk (rho: 0.49, 􀀀 0.47, p < 1e-270; MI: 0.17, 0.15), while age (DEMO_RIDAGEYR) exhibited a positive correlation (rho: 0.33, p < 1e-128; MI: 0.08). Total Correlation (32.68) confirmed significant multivariate interactions among predictors. These findings reinforce the model’s predictive strength, addressing Oka et al.’s recommendations and affirming the importance of anthropometric and demographic factors in osteoporosis risk assessment.

Palavras-chave

Osteoporosis risk prediction Machine learning Feature importance Statistical validation Spearman’s rho Kendall’s tau Mutual information Total correlation NHANES 2007-2014 Stacking ensemble