Status: Prioritization Needed
As a researcher, I want to DataShop to provide some report or extra data, so that I can grade online tests and have pre/post test data associated with my study.
The following notes come from an email thread between Bruce McLaren and Vincent Aleven. August 2010. There is also a related request from Ruth Wylie.
Provide a "last attempt" field" in transaction export
- Vincent suggested that DataShop simply put a new field in the transaction export to indicate which transaction is the last transaction for a given step. The researcher would still need to do some work in Excel to produce the actual grades for the students. But this seems relatively simple for DataShop to do.
- "Even if the DataShop is kind enough to implement the "test score" facility, I would hope it will also support the "last attempt" field, which would appear to be even easier to implement (since identifying the last attempt is a step in computing the test score) and provide great bang-for-the-buck. This field is still very useful for me and the projects that I have been involved with. For example, it is useful when the standard "test score" analysis is not quite what you need (e.g., when test items depend on each other in certain ways, or when you want more fine-grained measures of test performance, such as transfer v. reproduction - this is very very very typical and endorsed if not required by the PSLC theoretical framework). Further, given that we typically analyze test scores in Excel anyway, and there are typically so many ways to look at test scores, it seems important to offer flexibility (meaning, the "last attempt" field)." -- Vincent email 8/3/2010
- Related Request: Last Attempts - Test Item :: As a researcher (or educator), I want to see data for last attempts only so I can determine correctness on test items. -- Ruth Wylie, July 3, 2008
- A "last attempt" field and grading options would definitely be beneficial and something I would frequently use. It would also be very helpful to have DataShop (or this might be on the CTAT-side of things), automatically insert missed steps (per Alida's note on Aug 4, 2010) which might help some of the automatic grading issues? -- Ruth Wylie, email August 12, 2010
Provide two new fields in the student-step rollup
- DataShop can identify all problems where the tutor_flag field equals, 'test', 'pre-test' or 'post-test'. Then for each step in that problem, find the last transaction for that step and the outcome (CORRECT vs. INCORRECT) and then put two new columns in the student-step rollup. One column indicating whether the step is part of a test problem. One column indicating whether the student was right or wrong on the last transaction. -- Alida 09:52, 4 August 2010 (EDT)
- "I think Vincent's idea of adding a new field indicating whether an answer is the last one provided for a particular step is a great start. However, since it still requires (potentially error-prone) manual manipulation in Excel after exporting the data, why not take it a step further and also provide a DataShop capability to generate a "test score" (and a data file to give to teachers -- that would really help me) from a given set of selections?" -- Bruce email 8/3/2010
- Producing a grade that teachers could use seems difficult. How would DataShop know how many steps are part of each problem and how many problems are part of a test? If a student skips a step, and there are no logs for that step, then that student would have less steps than another student. -- Alida 09:47, 4 August 2010 (EDT)
- It is true that there are a variety of ways one can grade an online test with DataShop log data. However, I think there are a core of perhaps 4 or 5 different ways it might be done, none of which are particularly complex to implement. I'd suggest a DataShop interface that allows a few variations on grading that the researcher can select and that the DataShop implements. (And there would still be the possibility of simply using the extra field Vincent has requested, if none of the implemented approaches works.) -- Bruce 8/5/2010
- One huge advantage of providing several methods, available in one place (namely, the DataShop) is that the researcher could then rapidly experiment with the results of different grading approaches, ones they may not have tried otherwise, and then decide to grade their online tests differently than originally conceived. I doubt that this would be done with off-line, ad hoc grading, as things stand now, given the effort to implement different grading approaches for each project separately. I certainly would use such a facility.
- And here is another advantage: Recently, I have had several researchers ask me for the pre-post test scores for my stoich studies. Since I have always calculated these scores separately and off-line, they are not available in the DataShop. If the DataShop provided a grading mechanism, the scores for all experiments that have online tests could be directly and easily stored with the original data, thus facilitating future research on study data.
- Grading approaches I've used on stoich:
- Average all of the steps in the test (pre or post), taking the last submitted value for each step as the "final answer"
- Weighted average of all problems (1 problem = X steps), taking the last submitted value for each step as the "final answer"