TODO

1. Implement pure-birth (Yule) model - DONE
  - need to:
     - count parameters better in information theoretic calculations - DONE
        - remove 'NAs' from parameter counts
        - update stopping criterion (more models allowed) - DONE
     - pass model flavour to various functions (get.max.model.limit, make.lik.medusa.part,
       medusa.ml.fit.partition, medusa.ml.initial, medusa.ml.prefit, medusa.ml.update
       calculate.model.fit) - DONE

2. Implement mixed models (mix of BD and Yule processes) - DONE
  - need to:
     - calculate both flavours for both tips and virgin.nodes
        - put results in a common list 'prefit' - DONE
     - calculate both flavours for all splits, select best one for each - DONE
     - perform checks on all downstream functions - DONE
     - update summarizeTurboMEDUSA to consider flavour of model when plotting surfaces - DONE
        - at the moment assume only BD models are to be plotted - FIXED

3. Implement summary function for a results over a distribution of trees - DONE
  - just need to find a efficient way to parse the list of lists

4. Make surface plotting more general - DONE
  - fit plot to relevant region (i.e. not just [0,1] for both r and epsilon) - DONE
  - plots for Yule models - DONE

5. Pass in user fixed epsilon - DONE

6. Plot tip richnesses next to tree (needs to consider coordinates) - DONE, but placement a little ugly

7. Fix r (b) optimization interval for extreme clades - DONE

8. Complete documentation
  - especially for multiMedusa functions

9. Alter output such that only the final optimal model is saved to the results - DONE
  - this will require changing summary functions

10. Significant memory problem: intermediate results of lapply over nodes can be several 100s of MB for large trees (e.g. 2000 tips)
  - keep only AICc score and node ID during course of lapply, recalculate on best node after - DONE

11. Problem with conTree and multiMedusa trees possibly having a different tip.label ordering - DONE
  - reported by Laurie Sorenson (UCLA)
  - multiMedusa trees *should* have same ordering, as likely read in using a single translation table (e.g. BEAST)
    - not sure if this applies to reading in a file (e.g. phylip or nexus) without a translation table
      - CHECK THIS
  - for the first part, try reordering conTree tip labels (since multiMedusa should be fine)
  - now, explicitly manage tip label ordering upon input (suped-up version of ape:::.compressTipLabel())

12. Fix crashes due to poor initial starting parameters in some tree distributions - DONE
  - dealt with number of edges vs. total path lengths per partition

13. multiMedusaSummary dies when some shift types are not observed in the distribution (e.g. stem or node) - DONE

14. Error mapping richnesses to re-ladderized tree when plotting - DONE

15. Remove draconian requirement of labelling richness file columns with c("taxon", "n.taxa") - DONE

16. "subscript out of bounds" error when setting shiftCut="node" - DONE

17. Post installation/usage instructions - DONE

18. Tag results with MEDUSA version number - DONE

19. Fix multiMedusa summary error - DONE
  - involved trees with no (mapped) shifts

20. Fix general plotting.
  - text size and plot positioning should depend on tree size
  - remove species richness plot when all tips == 1 - DONE

21. Test for zero-length pendant edges - DONE
  - ape considers these binary and ultrametric

22. fix problem in threshold setting. using initial tree rather than pruned tree - DONE
  - will result in overly conservative models

23. fix error with shift magnitudes - DONE
  - rates are correct, but magnitudes are wrong
  - identified by James Horn

24. plot shift densities
  - need to make sure keep track of r vs. b

25. write figtree file with shift and rate annotations - DONE