WeBWorK Main Forum

WebWork for placement

WebWork for placement

by Eddie Fuller -
Number of replies: 4
Hi everyone,

I was wondering as I survey another mapleta crash during placement testing if anyone has used WebWork to do placement testing. I guess this is probably a stretch since the gateway module isn't designed for that per se, but any comments would be appreciated. Specifically,

*Are there issues that anyone knows of regarding copyright or other restrictions that would prevent us porting the MAA placement tests to the WebWork system?

*Has anyone tried this? In the case the the MAA tests can't be used, has anyone developed similar tests?

*Are there other folks who might be interested in such a set of exams if we were to develop them (provided some don't already exist)?

Looking at the forum history it looks like some people have done this so any experiential info would be great. We have lots of scaling data for the MAA suite in MapleTA, but it simply can't handle the load we put on it.

Thanks in advance,

In reply to Eddie Fuller

Re: WebWork for placement

by Robert Byerly -
I'll be interested in other people's experience about this too. We are in the process of porting our old on-line placement test (written for a now unsupported application called Onex) to WeBWorK. Our current test is I believed inspired by the MAA test suite but the problems were developed in-house. There is some data to indicate how well certain problems predict student success at our institution (Texas Tech University).

The port of the problems is incomplete, but mostly finished.

Since we have thousands of students taking the placement test each year, automating some of the book-keeping is a concern. I am developing some simple scripts for reporting scores to the Registrar's office automatically.

I'm planning to create a WeBWorK course with a sufficiently large number of dummy accounts (thanks to Gavin LaRose for suggesting this), and write a WeBWorK authentication module that will allow incoming students to authenticate through the University's login server, and which will automatically associate the student with an unused account if he/she hasn't taken the test before, and with the same account used before otherwise. If anyone has a better scheme I'd love to hear about it before I get too far into this!

Bob Byerly

In reply to Robert Byerly

Re: WebWork for placement

by Danny Glin -
Here at University of Calgary we run a small scale diagnostic test on WeBWorK. It is really more of a challenge exam for students who do not have the high school prerequisites for first year courses but feel they are prepared. We offer sessions of up to 30 people (as that's how many our computer lab holds) about half a dozen times a year. As it is not a large number of students, we currently manually create the accounts as they are needed.
We also run in-class quizzes for our first year calculus courses using the gateway module on WeBWorK. In this case we have set up a nightly dump of class list information from the registrar's office into WeBWorK to deal with registering students.
Reading about your situation reminded me of another option we were considering for importing students, which is as follows:
Have the student authenticate to the central authentication server (LDAP in our case). If a webwork user exists for that username, then log them in. If not, then create a new webwork user with that username. I know our LDAP directory contains full name and email address (among other information), so these could be queried from the server once authentication is successful and included when creating the new user.
The downside I see to this is that you will quickly end up with a huge amount of students in your classlist. One benefit of this is that with this model you would be able to retain data for all students more effectively.
Hope this helps...
In reply to Danny Glin

Re: WebWork for placement

by Daniel Smaltz -
At Worcester Polytechnic Institute we have been using WeBWorK for placement testing for the past two years. Three optional (!) placement tests are presented to every entering freshman to determine which of the introductory calculus classes they wish to take. We have a freshman class on the order of 800 students each year and have achieved about 60% participation on the placement tests, which are offered in the summer before students matriculate.

Our three tests are loosely based on MAA standards but were written and coded ourselves. One test covers pre-calculus and closely resembles the MAA Calculus Readiness Test, and our two other placement tests cover basic differential and integral calculus.

We have a script in place to create WeBWorK accounts for every entering freshman automatically upon creation of their school email address. The source-code details of this script aren't familiar to me, but if anyone is interested I can put them in contact with the appropriate person.

As a university requirement (I'm an undergrad), a partner and I completed a thorough study of the effectiveness of our WeBWorK placement tests. The report from that study is available online: http://www.wpi.edu/Pubs/E-project/Available/E-project-042408-132635/

Our advisor for that project, and the Dean of First Year Programs at our university, is Prof. Arthur Heinricher, who would be happy to discuss the use of WeBWorK for placement. You can reach him at heinrich@wpi.edu.

In reply to Eddie Fuller

Re: WebWork for placement

by Gavin LaRose -
Hi all,

We've used WeBWorK for our placement test at the University of Michigan for the past couple of years. We have about 7000 students who take the test in the course of the summer. Our test is based on the old pencil-and-paper test that we used to use before we moved on-line; when we first implemented the on-line version we did some evaluation that indicated that student scores on the on-line test were slightly higher than those on the pencil-and-paper test, but that students also did slightly better in the courses into which they were placed. The test is taken by students before they come to campus (theoretically, at least). I should add that the placements are strongly advisory, but don't actually have "teeth": a student can place into precalculus and then sign up for calculus III and the only thing that will stop her/him is her/his advisor. Our test only distinguishes betweeen placement into precalculus or calculus.

Specifically on the topic of the MAA tests, I think it would make sense to talk with the MAA about porting their tests to WeBWorK. I don't know what their attitude about the placement test content is, nor what their agreement with MapleSoft requires.

For anyone who is interested, some additional details about our placement testing setup follow.


Some details:
- Account creation: we finally have things in place so that by the time students need to be able to take the placement test they have their University login. We authenticate through the central login server, so we just create (approximately 6700) student logins in the placement course at the beginning of the summer, and those students who don't actually enroll just never show up with actual test data.
- Data recovery: I have a script that plows through the WeBWorK database to get test results. Because of the number of students we're working with, I store results in a GDBM database for every student I find so that on the next run I only have to look up new tests in the WeBWorK database. This is far, far faster than sorting through WeBWorK for every student. The script then creates a data file in the format needed by the people who actually generate placement results and send them do advising. The file(s) is(are) then dumped in a directory where they can log in to get the data.
- Incomplete tests: It's worth noting that if students don't hit "submit" before their time expires, they don't have a test score on record. It is, however, possible to get their last answers from the course answer_log. Because it involves sorting through a (potentially large: even after truncation in June, my answer log for this summer is over 180MB) this is not a fast operation. This is another reason I keep the cache of student results to avoid lookups.