Codes for "Predicting Childhood Obesity Risk: A Longitudinal Study of Children in Ireland"

These are the codes for the modelling process used in Predicting Childhood Obesity Risk: A Longitudinal Study of Children in Ireland. Researchers wishing to reproduce our results may apply for access to the GUI dataset through the Irish Social Science Data Archive (ISSDA) at www.ucd.ie/issda and to the MCS dataset through the UK Data Service at www.ukdataservice.ac.uk. Model Development ( Apparent and Internal Validation) Step 1: We pooled the candidate variables from the Growing Up in Ireland (GUI) study database. Step 2: RecodingCandidateVariables.R ----- Recode, create composite variables in the candidate variables in the GUI data Step 3: MICE_Strategies.R ----- Remove overlapping variables following expert knowledge and statistics strategies in the GUI dataset. Step 4: sample_size_calculation.R ----- Calculate the minimum required sample size for the GUI dataset. Step 5: MICE_GUI.R----- Multiple impute missing values and performed weighted lasso for variable selection in the GUI data Step 6: Development_process.R----- Develop and internally validate models based on AUC, calibration and decision curve analysis. External Validation Step 7: MCS_recoding.R ----- Recode external validation dataset to have the same format as the development dataset. Step 8: MICE_MCS.R ----- External validation sample size calculation and multiple imputation by MICE. Step 9: External_validation.R ----- Externally validate the developed models. Additional files referenced in the manuscript are as follows: Additional file 1: Detailed variable descriptions across the growing up in Ireland and millennium cohort study datasets. Additional file 2: A comparison between participants missing versus non-missing outcome data across the growing up in Ireland and millennium cohort study datasets. Additional file 3: A summary statistics for all included variables for non-obese and obese group for the growing up in Ireland dataset. Additional file 4: A summary statistics for all included variables for non-obese and obese group for millennium cohort study dataset. Additional file 5: A summary CSV file detailing the selection frequency of each predictor variable across 40 imputed datasets using various the modelling approaches. Additional file 6: Distribution of the predicted risks by outcome in millennium cohort study dataset. Additional file 7: The TRIPOD+AI checklist.