Include +I+R models

[iqtree]

Include +I+R models

[roblanf]

Just to add some evidence. Here's an analysis of a dataset from Endicott and Ho (2008), after merging partitions etc.

4 out of 6 final partitions have +I+R as the best model.

TN+F+I+R3: dloop,                                                                                                                  
HKY+F+I+R3: a61_c11_c21_c31_cytb1_n11_n31_n41_n4l1_n51,                                                                            
TPM3u+F+I: a62_c12_c22_c32_cytb2_n12_n22_n32_n42_n4l2_n52_n62,                                                                     
TIM3+F+I+R2: a63_a83_c13_c23_c33_cytb3_n13_n23_n33_n43_n4l3_n53,                                                               
HKY+F+I+R2: a81_a82_n21_12s_loops_16s_loops,                                                                                   
TN+F+G4: n61_n63;                                                                                                              

The reason it's a good model is that +R models can't have a rate category of zero, but +I+R (of course) do.

[roblanf]

@bqminh tagging you as discussed!

[thomaskf]

A new branch called mf2 has been created.
This request has been addressed in the branch mf2.

[thomaskf]

To download and compile the source codes in the branch mf2 :

git clone --recursive https://github.com/iqtree/iqtree2
cd iqtree2
git checkout mf2
mkdir build
cd build
cmake -DUSE_LSD2=ON ..
make -j