floriankrach/lms-model

LMS model for percentile curves

LMS-model

Reference

Article Title: Reference values of body composition parameters and visceral adipose tissue (VAT) by DXA in adults aged 18–81 years—results from the LEAD cohort

DOI: 10.1038/s41430-020-0596-5, 2019EJCN0971

Link: https://www.nature.com/articles/s41430-020-0596-5

Citation: Ofenheimer, A., Breyer-Kohansal, R., Hartl, S. et al. Reference values of body composition parameters and visceral adipose tissue (VAT) by DXA in adults aged 18–81 years—results from the LEAD cohort. Eur J Clin Nutr (2020).

Description

This is a repository with the entire code used to create the LMS models, plots and statistical analysis for the paper referenced above. The code should be easy to understand and use for other projects including LMS model generation for body composition data of large cohorts, in particular, to generate age dependent (in a continuous manner) percentile curves which can be used as reference values.

The code is written in R.

Data & Applications

While the raw input data is not publicly available, the generated LMS model output data (LMS values and percentile curves for children and adults) can be found here. It is part of the Scientific LMS Z-Score App, a python application to compute z-scores based on our generated reference values.

A more easily accessible version of this is the corresponding WebApp.

Citation

If you find this code useful or if you use it for your own work, please cite the paper referenced above.

lms-model's Issues

Data files?

Hi @FlorianKrach,
Thanks for posting the study's code on github! Would you mind uploading the input and output datafiles from this script? Specicially these files:

data_file_name = 'data_reference_values_v17.xlsx' # the name of the data file
save_output_file_name = 'data/auto_overview.xlsx'

I realize the reference values should be the same as the data in the tables in the paper but having it in excel format along with your percentiles would make using the data far easier.

Before I found this repo I had already extracted the values of the percentiles here but getting the actual data from you would be far better.