select a bit more

vignettes/selection.qmd

@@ -112,5 +112,4 @@ The function `chooseDist()` chooses the "best" distribution according to a GAIC
 ## Select appropriate variables  {#sec-Selectappropriatevariables}
 
 The methodology of "which explanatory variable is needed and for which parameters" depends on the scope of modelling,
-while special attention has to be taken for **non-linear** relationships and for **interaction**.   If the scope of the model is to provide  good interpretation then the additive smooth structure of `gamlss2`is adequate because can cope well with non-linearities and relatively  well with interactions. Note that interactions in an additive structure have to specified explicitly while in some machine learning algorithms like neural network they come part of the model. Machine learning algorithm like random forest and neural networks are more difficult to interpreter.        
-
+while special attention has to be taken for **non-linear** relationships and for **interaction**.   If the scope of the model is to provide  good interpretation then the additive smooth structure of `gamlss2`is adequate because can cope well with non-linearities and relatively  well with interactions. Note that interactions in an additive structure have to specified explicitly while in some machine learning algorithms like neural network they come as part of the model. Machine learning algorithm like random forest and neural networks are more difficult to interpreter therefore  more suited for prediction purposes. Here we will discuss the the **step-wise** methodology while boosting is discussed in a different  vignette?.