Student Assessment

This topic was discussed during the Intro Stats group meeting on Dec 1, 2008.

Content

Homeworks, Exams, Weighting, Plagiarism, iClickers/Quizes, Unclassified notes.

iClicker Questions for STAT 200 shared by Eugenia.

Homeworks

  • Jonathan: Many small tasks is better than one big.
  • Jonathan: People procrastinate. Deadlines spread along the term make it easier for everyone. See book Predictably Irrational for a description of experiment where students with multiple deadlines did better than those who had to hand in everything in December. Bruce agrees with this and also says that there is some research available on this topic: it's important to make students work throughout the term.
  • Jonathan and Mike W think that homeworks' primary role is just to give students an idea of what they are supposed to know and give them the opportunity to work throughout the course. It's not really assessment. They only assign 10% of the final grade to the homework assignments.
  • Mike W: Regular assignments allow the instructor to track progress of students.
  • Everyone accepts assignments in paper format at the moment. It makes it easier to grade it in front of TV. Jonathan accepts them online for one MBA course only
  • Nancy: In Stats department we ask students to submit computer code together with their assignment. However, in other disciplines, they won't always have the source code because of the point-and-click nature of the software.
  • Jonathan would allow students who have difficulties with computing to work with someone but they would need to indicate who they work with.
  • Steve tells their students that they can work in groups but have to write themselves. Students still cheat.
  • Nancy: We supplement assignments with lab activities.
  • Nancy: Want to give students more in-depth problems on assignment. This sort of contradicts to the "many small ones" principle but can be solved by giving incremental problems where the problems becomes more and more complex but deadlines are still regular.
  • Nancy: Never have students hand anything in at the end of the class—they will not pay attention to your teaching.
  • How many homework assignments to we give:
    • BIO 300: 10 assignments (weekly), 1–4 questions each
    • STAT 2xx: 3 assignments, 2–5 hours each
    • COMM 231: 6 shorter assignments

Projects

  • Projects are great but unfortunately they cannot be used in large courses because of the grading burden: a lot of content and difficult to have a fast and fair grading scheme.
  • Group projects are bad for assessment because one person in group usually does all the work.
  • Jonathan has tried giving a data analysis problems in homeworks but then would only grade the format of students' reports but not the actual content.

Exams

  • Statistical problems and exam questions can be easily shared across disciplines and universities. Lets share exam questions but since it's a sensitive material we won't post it on the web. Email Eugenia to see what's available or to share your own collection of questions.
  • Jonathan's view of exams:
    • Include some Multiple Choice, some Short Answer and very few Long Answer questions. Space given for the answer on exam sheets is indicative of how long the answer is expected to be.
    • Making questions semi-independent is important and challenging.
    • No questions that require only to reproduce a fact/formula from textbook. They should be: know something and be able to use it in a simple derivation. Example: what is the ratio of regression coefficients "Y on X" over "X on Y"?
    • Only standard calculator is required for the exam.
  • On midterms:
    • STAT 2xx have two midterms to reduce the weight on the Final.
    • Jonathan thinks that midterms are very distracting for students as they start studying for exams instead of learning. But what about "many is better than few" principle for homeworks? Does it or does it not apply here?
    • Mike W always explains fundamental things crucial for understanding the course as a whole during the lecture before exams and students pay a lot of attention. He also clearly tells them what they are supposed to know and be able to do in the exam.
    • Rob allows students to drop the midterm grade and put the weight on the final. But students don't know about it before the exam and have one week after they get their grade to decide whether they want to keep it on bank on the Final. Steve and Mike W do something similar.

Weights for the total mark

CourseFinalMidterm(s)LabsHomeworkOther
COMM 29155%35%10%
BIO 30050%30%10%10%
STAT 20045%25%8%15%(3x5%)iClickers: 5%
Online Survey (2): 2%
STAT 20345%50%(2x25%)5% (tutorial quizes)
FRST 23145%30%(see note)25%
PSYC 21840%42%(20%+22%)15%Participation: 3%
  • In FRST 231 students are allowed to drop Midterm grade and put the weight on the Final but they don't know it prior to the exam and then have a week to make the decision.
  • Carrie mentions that in PSYC the Final's weight must be less than 40%. Carrie original put 25% on lab assignments (in PSYC 217?) but now thinks it's too much because of cheating.
  • Jonathan mentions that a lot of research has been done on weighting schemes that shows that they are mostly irrelevant.

Plagiarism/Cheating

  • Mike W and Jonathan think that homework is there for the students' benefit and they may cheat if they don't want to take advantage of it.
  • Acknowledged help and collaboration is usually allowed as long as submissions carry some individual work.
  • In Stats department we make them include computer code that makes identifying plagiarism a little easier.
  • Carrie has a piece of software that can detect cheating in multiple choice exam based on the seating plan.
  • Students taking STAT 2xx courses have been caught cheating with Clickers (clicking for themselves and for a buddy).
  • Overall it seems like cheating on all forms of homework is very prevalent and brings a lot of frustration especially to junior instructors.

iClickers and In-Class Quizes

  • Clickers have been very successful in intro stats courses taught by the Stats department. Here are some of Bruce's points:
    • We have used clicker since Fall 2007 and we are happy.
    • Clickers show you the truth—many students do not understand your explanations even if you think they are crystal clear.
    • It's a good way to make students participate in class. With clickers they actually do answer questions that would produce just a dead silence otherwise.
    • Very little weight in the final grade: 3%. Half of it is just for participation and half for giving correct answers.
    • Have caught them cheating.
    • Ask 3–4 clicker questions per class.
    • Start with questions about the content of the previous class.
    • Ask questions that they can get wrong. Let students see that many people get it wrong and they will be less ashamed of showing their opinion.
  • The software that collects data and summarizes it for grading: iClicker, iGrader.
  • Jonathan did low-tech clicker sessions to encourage participation and to convince students that questions can be challenging and that anyone can answer them wrong. He would come up with a couple of True/False statements that are easy to get wrong (common mistakes), ask the class to think about them and would ask if anyone is 100% confident in their answer. A few people would raise there hands, the instructor would choose one of them and ask for their answer which would be wrong with a high probability. Then explain the correct answer to the class. Students then can see that they are not the only one who sometimes misundertand material and make mistakes so they will be less afraid of participating in class (and running the risk of being wrong).
  • Jonathan also mentions that iClickers cannot be used for student assessments in very large courses he oversees because of the difficulty of standardizing the result from different sections taught by different instructors.
  • An agreement has been reached to share iClicker questions with each other. Here are some.

Extra Discussion (Jan 28, 2009). MW brought up the issue of context for iClicker questions suggesting that just questions themselves have limited value if not accompanied by the where in the curriculum they were used. Inspired by this the following points have been made:

  • EY: clickers are used (1) to test if students understand concepts after they are explained; (2) to review material from previous lectures at the end of the class.
  • NH asks clicker questions about old concepts when they arise in new material (to segue).
  • BD: clickers give feedback to the teacher. According to Carl Wieman's method, ideal students should read about lecture material at home before class, and then the professor will teach the class based on clicker responses to cover topics that are the least clear to students.
  • CC: clickers make learning more interactive. Also, Carie would teach basic material first and then ask more advanced questions to see if students can apply their new knowledge: if they get it right then there's no need to spend too much time explaining it in class.
  • JB uses counter-intuitive examples in his stories in lieu of clickers. Students get intrigued/excited about the mystery and they also feel superior because they know something special.
  • (update July 28) EY: after a clicker question, explain the theory behind the question before giving the answer — this way you will have students' attention as they are interested to know if they got it right.
  • (update July 28) Clickers increase attendance because students come to class to get the attendance points, which is good. But it also means that you will have students in class who are not interested in what you teach which is (1) distracting to everyone; (2) will distort your feedback from clickers. See discusion on July 28, 2009 for more thoughts on teaching stats to large classes.

Also, should we put more iClicker questions online? Password protected? No definite decision has been made.

Unclassified notes

  • Online exams. Bruce mentions that there are systems that allow instructors to produce customized/randomize exam questions that can be used for online/in-lab testing. Rumours are that Vista can do it.
  • Rob recommends going to your own labs (conducted by a TA) to get instant feedback on how well students are following the material.
  • Jonathan states that students' performance has gone down in recent years as he was giving fewer assignments and putting more notes online. After a cohort of students was given the same test as another group of students 5 years ago, they scored 7% less on average.
  • COMM 291 has no labs but students are taught how to use MS Excel in lectures.

a place of mind, The University of British Columbia

Department of Statistics

E-mail: slate@stat.ubc.ca

Emergency Procedures | Accessibility | Contact UBC | © Copyright The University of British Columbia