Warning messages from the model

[stephaniereinders]

Several things can go wrong with the pipeline.

Too many outliers

If "too many" graphs are assigned to the outlier cluster for the cluster template docs, the model training docs, or the questioned docs, the code should return a warning.

To Do:

• Ask Alicia and Danica how we should define too many outliers
• Add warning to make_clustering_templates()
• Add warning to get_clusterassignment()

Empty clusters

examples_empty_cluster.pdf
In the attached example the cluster template has 5 clusters. None of the graphs from the model training data are assigned to cluster 5. I asked Danica how we should handle this and she said that handwriter should return an error suggesting that the user create a new template with fewer clusters.

Need to decide which function(s) should throw the error. Maybe the format_model_data() function? Then format_questioned_data() should throw an error if the questioned data doesn't use all of the clusters.

Model docs and questioned docs use different clusters

I didn't record an example, but occasionally the questioned docs will use different clusters from the model docs and this throws an error in analyze_questioned_docs().

Need to decide how to handle this. Maybe format_quesitoned_data() should throw an error if it uses different clusters from the model docs?