Mixing different data types in GGM - preprocessing ? about bootnet

[SachaEpskamp]

Hi Wivine,

Centering and scaling is typically not needed for models in bootnet as the models show only standardized effects. As for what transformations may be appropriate that is very hard to say and really depends on the data. The Spearman correlations / rank transformation generally work well if data are not normal, but for fine grained continuous variables perhaps the NPN is more appropriate. For count variables it would be better to use mgm (which is supported in bootnet) together with the Poisson type for those variables.

Best,
Sacha