C.1 Bugs Fixed in Version 5.1.0 of ChemTree

  • If no descriptors at all have either been imported or generated from the imported compounds, ChemTree will show a nice message saying so, rather than crashing.
  • If the only field in the “footer” of a compound is one that is needed for the compound name, ChemTree will properly allow you to import both your compound and your descriptor/potency file, rather than giving you a strange error message.
  • The error message that occurs for an SD file error will continue to work properly, even if it must be shown many times.
  • Resolved performance issue with Mac OS X progress dialogs by switching to a more plain GUI style.
  • Long row names no longer overwrite other text in plots.
  • Quickly deleting multiple navigator nodes caused a crash.
  • Saving a large tree image to bitmap crashed the program by exceeding system resources.
  • Focusing upon certain viewers did not always cause the correct navigator node to highlight.
  • Fixed long-standing bug where the p-value of the chi-squared test was being multiplied by two. This impacts tree splits when there is a binary response and a binary predictor, or where the predictor is categorical, integer, or continuous and has two unique values. It used to be possible to get p-values as large as 2.0.
  • The option to allow or not allow Navigator node deletion now works (and does not crash the program).
  • Problems with joining spreadsheets by sorting, when subsets were involved, have been fixed.
  • Better checking for possible singularities in the data involved with finding splits is now done.
  • Proper labels for the P-Values are used for prediction spreadsheets. (Previously, “P” meant what should be called adjusted P, and “adjusted P” was used for the Bonferroni P.)
  • The odds ratio display within the categorical split display (for two-category predictions of binary responses) has been fixed to handle zero and infinite ratios, thus making it work better for multivariate responses.
  • The proper navigator node is highlighted when a new non-tabbed viewer is first created. (This has always worked properly for tabbed viewers.)
  • Fixed bug in distance matrix plots that crashed program when the cancel button was clicked while calculating.
  • Fixed bug in importing files with missing data in the first row that causes the row to be classified incorrectly.
  • When viewing a histogram of a 5 variables, because of the length of the variable names, it runs into the N1 (node name).
  • Fix bug in the project navigator that caused the program to crash when selecting "delete" from the right click menu and no items in the navigator are selected.