An empirical study using permutation-based resampling in meta-regression

Table 6 Results of final stepwise models relative to number of trials per variable and statistical heterogeneity ranked by number of trials per variable in the final models

Model	TPV	I²	Large sample P value for the final model	Permutation P value for final model
Allium sativum TG: model 1	22	98.1%	0.033	0.047
Hypericum perforatum HAM-D: model 1	22	81%	0.035	0.041
Allium sativum HDL: model 2	9.5	45.8%	0.003	0.003
Allium sativum HDL: model 2	9.5	45.8%	< 0.001	< 0.001
Hypericum perforatum HAM-D: model 2	8	79%	0.021	0.022
Hypericum perforatum HAM-D: model 2	8	79%	0.041	0.045
Allium sativum TG: model 2	5.7	87%	0.031	0.036
Allium sativum TG: model 2	5.7	87%	0.001	0.003
Allium sativum TG: model 2	5.7	87%	0.008	0.016

This table represents the final models for standard large sample techniques, the statistical heterogeneity in the final model, and number of trials included in the final model and the P values for both the standard large sample techniques compared to the permutation-based resampling P values. Model 1 refers to the stepwise model for methodological covariates. Model 2 refers to the stepwise model for the participant level and intervention level covariates.
HAM-D = Hamilton Depression Scale score; TG = triglyceride; TPV = number of trials per variable (TPV) in each model.

ISSN: 2046-4053