The comminterpretation table has fields classfit and classconfidence that are sorta supposed to conform to a controlled list, but this is not enforced anywhere, and the database has various unexpected values. It would be nice to translate these into allowed values (and then decide whether to implement constraints in the API or DB to keep these from drifting again in the future).
- classfit: The data dictionary (here) says we should have only 5 string values, but actual data has 6 ("High" doesn't belong), plus numeric values 1-5. The numeric values likely correspond worst to best ordering -- but occasionally 1 can be best? May need to manually review plots with these values to verify the ordering. "High" is likely meant to be a confidence indicator (see below)?
select classfit, count(*) from comminterpretation group by classfit order by classfit;
/*
classfit | count
---------------------------------+-------
1 | 109
2 | 1147
3 | 4384
4 | 5048
5 | 4081
Absolutely correct | 24086
Absolutely wrong | 2
Good answer | 3660
High | 2
Reasonable or acceptable answer | 518
Understandable but wrong | 2718
| 57196
*/
- classconfidence: Data dictionary (here) says we should have only 3 string values (High, Medium, Low), but actual data has various abbreviations of these, plus four records with the number 1 (perhaps corresponding to fit "high" in the comm fit field?).
select classconfidence, count(*) from comminterpretation group by classconfidence order by classconfidence;
/*
classconfidence | count
-----------------+-------
1 | 4
H | 3249
High | 5385
L | 170
Low | 327
M | 1539
Med | 132
Medium | 845
h | 1
high | 658
low | 333
med | 381
medium | 14
| 89913
*/
The
comminterpretationtable has fieldsclassfitandclassconfidencethat are sorta supposed to conform to a controlled list, but this is not enforced anywhere, and the database has various unexpected values. It would be nice to translate these into allowed values (and then decide whether to implement constraints in the API or DB to keep these from drifting again in the future).