- 4/24/2015 3:57:10 PM
You have a point. In a sense the TA defines the grade, so it is what it is. ("TA ___ gave Student ___ a score of _____")
Let's change the analogy for a moment and think about mortgage application records. One of the entries in each record is "Income". Now, in our example, one loan officer always completes the applications with Net Income and another officer uses Gross Income.
If the business rules are crystal clear on the definition of this field, then one loan officer is just plain wrong. And that's not a problem with the business rules.
However, if the business rules are fuzzy - then this is a problem with the business rules.
- by erked, Prospector
- 4/24/2015 12:50:00 PM
In general, I'd say the problem is one of analysis and inference, not business rules. Recording the facts ("TA ___ gave Student ___ a score of _____") has no bearing on how the data are used to derive grades (which are not recorded facts, but essentially derivations using rules not known to the DBMS). A curve or non-curve policy, and the weight that the TA score is going to have on each student's grade, and adjustments made to compensate for differences in aggregate TA scoring... those are all a different realm, at least to my way of thinking.
- 2/17/2015 5:07:39 PM
No, my posts are about general fundamentals, not specific situations.
The next post is about what analysts who may not have designed the database should be looking for to try to overcome the lack of rule documentation and what questions to ask the designers to figure out what analysis makes sense, how to interpret the results and what qualifications they need to add in case they can't.
- 2/17/2015 4:23:14 PM
If your next post is about the Analyst curving the TA grades in the DBMS, I'll be interested. Data Analysts are too often assumed to be responsible for data integrity of data they had no hand in creating.
We didn't make the mess, but we're supposed to clean it up.
- 2/17/2015 3:40:09 PM
I did explain: Had the prof/univ imposed an explicit rule on grading, then yes, but it would have been difficult to enforce IN THE DATABASE by a DBMS.
It is easier for the prof to check the distribution of grades post-facto and adjust.
From a database perspective I'm concerned with business rules that are enforced in the database by the DBMS via integrity constraints (see next post).
- 2/17/2015 3:14:21 PM
not really business rules because they are not explicit
Please explain this further.
If the college has an explicit policy that students grades should not depend on which TA taught their lab, does that make this a business rules topic? Or if the prof does curve the grades?
- 2/17/2015 2:26:51 PM
You have no idea how close to home this hits. I used to be a TA.
These are not really business rules because they are not explicit--just the behavior of the TA's.
Had the professor establised a common and explicit grading policy for both TA's, perhaps, but it would have been non-trivial to enforce, as the rules would be about the distribution of grades.
Be that as it may, the prof can readily find the diff between the 2 distributions and normalize the grades. It's called grading on the curve and I did not like it.
- 2/17/2015 1:57:18 PM
As a real-world example related to the topic of undocumented business rules -
One college course has two labs. The labs are taught by different teaching assitants. Students are randomly assigned to the labs, so there are good students and not-so-good students in each lab.
One TA gives the average students a 90/100 and the best quarter of the students get 100/100 on every lab. The other TA expects much more. Regardless of heroic effort, his grades average 70/100 and no student ever gets better than 85/100.
Of course, if the professor understands the business rules for the lab grades table, he will take some action to remove the TA factor before turning in the semester grades.
And if not... students who had the wrong TA will see their overall grade drop. Anyone looking at the data table after the fact is not likely to find this problem.
- 2/16/2015 5:15:25 PM
We all know that but we use ordinal scales all the time and ASSUME comparability. How you interpret the results under this assumption is a completely separate and orthogonal issue.
The point is my argument holds for ANY AND ALL DATA and has nothing to do with the comparability of ratings.
- by tomsg, Data Doctor