gms | German Medical Science

In order to obtain more precise information, larger and larger studies with more and more measurements are being carried out. Suitable statistical methods are needed to analyze such a large amount of data. A special case is the analysis of repeated measurements. For example, in the large multi-center German National Cohort [1], also known as NAKO Gesundheitsstudie, the 24-hour food list [2] is repeatedly applied to estimate the individuals' consumption probability. For a large number of foods, only the question is asked whether or not they were consumed the day before. A first choice for modeling such repeatedly measured binary outcome variables is the logistic mixed model with random effects. The drawback is that this model has a very long run-time for large sample sizes. Therefore, as an alternative, the Multiple Source Method (MSM) [3] was revised to estimate the probability of the occurrence of a repeatedly measured binary variable for the application in large studies. The MSM uses a shrinkage technique of the residuals from ordinary logistic regression. These residuals are mathematically reduced to sums and thus no complex maximization algorithms are necessary, which are required when applying mixed models. Therefore, the MSM allows a fast calculation and can be applied in large studies. The result of a simulation study is used to represent the properties of the revised multiple source method.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.