COGNITIVE TESTING TOPICS

Well, it is February 21, 1996 ... and we are just beginning to discuss topics related to developing cognitive tests ... for example like for use in classes to assess student achievement. I first handed out a list of possible topics we would be discussing ... and suggested that they prioritize the items ... because we will not in all liklihood be able to discuss all. If you have anythoughts out there in cyberland as to the most important ones ... send me a note ... Roberts. Here is the list:
  1. Reliability/validity
  2. Normalizing distributions
  3. Simple concepts of test equating
  4. Cheating on tests and detection/reducing
  5. Absolute or relative grading standards
  6. Correction for guessing formula
  7. Speededness in tests
  8. Model of test settings
  9. Primer on factor analysis
  10. Grading on the curve
  11. Item writing rules
  12. Alternative ways to give/respond to/score MC tests
  13. Item types
  14. Problems with the None of the Above option in MC tests
  15. Test construction model ...
  16. Item analysis with ITEMAN
  17. PSU policy manual on exams/grading

Test/Scale Construction Model

We spent some time on what I called the test or scale construction model. The model has 3 overall stages or major side headings: PRE MEAS, MEAS, and POST MEAS. Each is discussed briefly below.


TEST/SCALE CONSTRUCTION MODEL



MEASUREMENT STAGE        NAME OF STEP      ACTION



PREMEAS                    Objectives      What is it you want to 

                                           instruct on? Attitudes

                                           to change?



PREMEAS                    Instruction     Provide actual instruction



MEAS                       Table of Specs   2 dimensional table of

                                            topics/cog levels



MEAS                      Item compilation   Construct/gather/ items

                                             appropriate to TOS



MEAS                      Administration      Give the darn thing!



MEAS                      Analysis I          Make test scores and

                                              show group summary



POSTMEAS                  Evaluation         Item analysis and 

                                              other info ... make

                                             final judgement as to

                                             "worth" of this test

--------------------

Start the whole process over again ... have fun!

The pre measurement phase includes bothy 1) defining objectives, and 2) allowing for instruction/activities/experience. As for 1, it is assumed that a typical course of instruction will necessarily need to have some up front objectives ... both for what the instructor is planning on doing and (more important I guess) for what the students are expected to accomplish. I mentioned in class today that back in the late 60s and early 70s .. there was a wave of interest in "behavioral objectives" ... (3 parts: task, setting, behavior criterion) ... and this swept the nation. But ... and it took about 10 years to realize this ... that if one spent all ones time (and you could) on developing behavioral objectives ... one did NOT have any leftover time to provide instruction. So ... there was a gradual diminution of the stress on these ... but some overall listing of objectives is still a good idea. As for 2, after the objectives have been defined ... listed ... there needs to be time for students to have opportunities to accomplish what was laid out in front of them as THEIR tasks. So ... instruction occurrs ... activities for practive are given ... ie, one experiences things to help achieve the pre defined objectives.

The measurement stage is where you begin to get down to the brass tacks of planning and developing and giving an ACTUAL test. Here, there are 4 subheadings: 3) table of specifications, 4) gather/construct items, 5) test administration, and 6) analysis I. Again, each is mentioned briefly below.

For 3, it is helpful to develop what is called a table of specifications or TOS. This is nothing more than a 2 dimensional table ... where you have topics on the top (item rules, item analysis, etc.) and several levels of thinking/processing on the side (facts, skills, application, etc.). Then, after making the TOS ... one tries to assign percentage weights to each cell in the table accoriding to how much time/emphasis was spent in that area. NOTE: THESE PERCENTAGES SHOULD not REFLECT WHAT YOU wanted TO DO ... BUT RATHER WHAT YOU did! With the TOS, one can then get a feel for how many items you will have to find or construct to fit into each cell. Then, for 4 for example, if you were making a 50 item MC test and, one of the cells in the able was valued at 10%, then you would need approximately 5 items to adequately cover that area ... to have a balanced representation on your test, given the number of items you will be using. Of course, here is where the concept of an ITEM POOL IS VERY USEFUL .... where you collect over time, items that have been tried out and fit various categories so that come test building time ... you are SELECTING items ... rather than worrying about will you have time to CONSTRUCT items. 5 is test administration and that is rather self explanatory. But 6 ... Analysis I ... means the basic scoring you use to assign scores to students ... which is actually the end of the MEASUREMENT phase (since measurement is the assignment OF the numbers ... and now you have done that). You also might do some descriptive stats to show what happened on the test.

The last stage is the POST MEASUREMENT which in a nutshell, is the final analysis part and coming to grips with how WELL the students accomplished the objectives or to make final determinations as to how well your TEST worked. Here we might do some item analysis ... and make judgements about keeping some of the items in your item pool ... but possibly revising and/or discarding some that don't work too well.

Paired Comparisons Minitab Macro File
I also spent some time going over a Minitab macro file I whipped up on Tuesday ... that would actually do the kind of analysis that was mentioned in the previous section on Paired Comparisons ... that would take the input data of preferences and then convert them into final scale values. I tried to link the steps in the Minitab macro to the actual steps you need to go through in the analysis. In fact, when I finished this ... I discovered a mistake in one of th calculations I had reported ... but correcting that at this point is rather inconsequential.

If you want to look at the macro file ... follow this link ... Minitab Macros ... down the list of macros to #37 ... called paired comparisons.

Friday is the deadline for students to submit their topics for the classroom/cognitive tests they will be working on ... keep posted!

FIRST CUT OF TOPIC SELECTIONS/ FRIDAY FEB 23
Here is a tentative list of the topics given by the students ... today. I might work with them a bit to hone in on something a bit more specific.
  1. Inga ... Test on Learning disabilities, mental retardation, attention deficit disorder
  2. Hsiu-Ping ... Build a test for Management 301/304 ... Basic Management Concepts
  3. Juhu ... 3 possibilities: Non verbal test for children, preshool teacher readiness, preschool teacher efficacy
  4. Boris ... A test related to Health Ed 450 ... Worksite health promotion programs
  5. Helda ... A test on psychoactive drugs ... (hurry up Helda ... maybe the CLASS will need to take it!
  6. Rick ... A test about discipline of health education; ie, health and wellness
  7. Tim .... Test about information on learning, cognition, and memory ... don't forget!

TABLE OF SPECIFICATIONS

The main part of the class today focused on what is a table of specifications ... and how it can help one make better tests. Look at the following.


TABLE OF SPECIFICATIONS



         |               TOPICS

Cog Lev  |Item Rules (20)  Item Anal (20)  Rel (30)  Val (30)

-------------------------------------------------------------

Appl 20   |       4              4            6         6

          |------------------------------------------------- 

Skills 30 |       6              6            9         9

          |-------------------------------------------------

Facts  50 |      10             10           15        15

------------------------------------------------------------

Look at this very simple table of specifications. Assume that you are planning a test that covers units on Item Rules, Item Analysis, Reliability, and Validity ... and you have SPENT approximately and EMPHASIZED approximately at the percentages listed: 20,20, 30, and 30. Also look at the side heading where I have indicated several different cognitive processing levels that you have focused on (note: Bloom's taxonomy and others can be used there) and have ESTIMATED that you have spent about 50% on facts, 30% on Skills, and the remaining 20% on Application.

If you then crossmultiply a topic value times a processing value ... you will get the approximate emphasis IN THAT CELL ... so that applications for item rules is about 20% of 20% or about 4% overall. This then gives you are rather easy way to figure out about how many ITEMS you will need for the test ... and this helps you to make a BALANCED test in terms of topic coverage and cognitive processing coverage.

Now ... it this were a 100 item MC test ... the values above would also be the number of items you would need ... but say you wanted a 50 item MC test .... then we would simply need HALF of the value above ... the # of FACTS items would be (for example) accordingly ... 5, 5, 7/8 , and 7/8.

We then had some discussion about the following issues: creating an ITEM BANK to allow you to SELECT items for the test ... rather than having to constuct them each time, and how the number of cells you need to cover says something about sampling that is possible and, perhaps the types of items that you will be using. For example, what if you wanted to use essay items ....? In the first place, essay items are poor choices for assessing whether students know the facts ... so, does this mean you would NOT test in those cells on this test? If not ... content coverage suffers. Even if you decide on MC items ... you can see that some cells will have rather few items ... some only 2 perhaps ... and that does leave alot of room for sampling ... and perhaps MISSING the mark.

I also tried to say that far too many times ... if one were to compare what they actually DID in the class with your ESTIMATES of how much time you spent on various things .. you would be amazed at the discrepancies that can occur. Thus ... the table of specifications can assist in making a balanced test ... one that fairly covers the material in about the same proportions that were emphasized in the course. If that is not done ... how is the student to possibly know how to study for the test?


And .... everyone have a nice weekend! See you next week! Send me a message if you have some comments or want to ask a question ... Roberts
ITEM WRITING RULES ... follow this link Item Rules