#57 [mfilter] Feature Request: stepwise regression feature

SiLeBAT / FSK-Lab

FSK Lab

https://foodrisklabs.bfr.bund.de/fsk-lab/

GNU General Public License v3.0

2 stars 4 forks source link

#57 [mfilter] Feature Request: stepwise regression feature #240

Open miguelalba opened 6 years ago

miguelalba commented 6 years ago

Currently users have to define each formula explicitely to use the parameter fitting node. Nevertheless it is common habit, that especially for secondary models different formulas are tested and the one that fits the data best is then used. Accomplishing this with PMM-Lab right now is not feasible, as e.g. in case of a polynomial of 2nd order with 3 independent variables considering all possible interactions would mean to type in more than 20 different formulas. In R there are libraries that perform forward and backward stepwise regressions which one could look at.

miguelalba commented 6 years ago

[cthoens] - milestone: 1.0 --> 2.0

miguelalba commented 6 years ago

[cthoens] - assigned_to: Christian Thöns --> Jörgen Brandt

miguelalba commented 6 years ago

[cthoens] Assigned to joergen7

miguelalba commented 6 years ago

[joergen7] A stepwise regression using the R snippet, has been accomplished. Integration in PMM-Lab would represent a major task.

miguelalba commented 6 years ago

[mfilter] - assigned_to: Jörgen Brandt --> Christian Thöns

Priority: Normal --> Low

miguelalba commented 6 years ago

[mfilter] It is already possible to use the KNIME R nodes to perform stepwise regression. To do so one has to apply the XML2Table nodes to supply the R nodes with the relevant input tables and can then use the R "stats" library with its function "step" to perform stepwise regression. Therefore I reduce the priority to implement this as a PMM-Lab feature to "low".