Generalized Linear Models are constructed to account for situations of varying complexity, including: dependence of biochar properties on single or multiple predictor variables, where dependence on multiple variables can have additive and/or interactive effects; non-linear relation between the response and predictors; and non-Gaussian data distributions.
Because the true number of clusters for experimental datasets is unknown and cluster size estimation methods such as the gap statistic (Tibshirani et al., 2001) or silhouette width (Kaufman and Rousseeuw, 2005) are often unreliable, the EMVC algorithm is executed on multiple variable clusterings where the number of clusters varies over a specified range.
In cases where multiple variables affect the outputs, RSM is an efficient procedure for investigating the relationships between the dependant (response) and the independent variables [8].
The thinking is linear, where multiple variables contribute differentially in measurable ways to determine a measurable outcome.
In multivariate imputation, where multiple variables have missing data, there are two approaches that are available in standard statistical software.
Where multiple variables were included in a multiple regression block, each variable was evaluated by examining the significance of the variable coefficient.
This approach can be extended to cases in which multiple variables have missing values, where the sample is split into cases with a given variable observed, or missing.
Such direct analyses are useful during exploratory and descriptive studies, but not for more complex models of communication processes where multiple variables may predict a single outcome.
It is unclear how this method would extend to missingness in multiple variables, except for the case where only one of the variables is MNAR [ 18].
We also include results of probabilistic sensitivity analyses (PSA) where inputs for multiple variables are sampled from distributions for multiple cohorts.
