CUTLO= cut off responses with low expectations = 0 logits, no |
Use this if guessing or response sets are evident. CUTLO= cuts off the bottom right-hand corner of the Scalogram in Table 22.
Eliminates (cuts off) observations where examinee ability measure is CUTLO= - logits or more lower than item difficulty measure, so that the examinee has a low probability of success. The elimination of off-target responses takes place after PROX has converged. After elimination, PROX is restarted, followed by JMLE estimation and point-measure and fit calculation using only the reduced set of responses. This may mean that the original score-based ordering is changed.
CUTLO= is equivalent to Waller's procedure in Waller, M.I. (1976) "Estimating Parameters in the Rasch Model: Removing the Effects of Random Guessing", Report No. ETS-RB-76-0, Educational Testing Service, Princeton, N.J. https://files.eric.ed.gov/fulltext/ED120261.pdf
Usually with CUTLO= and CUTHI=, misfitting items aren't deleted - but miskeys etc. must be corrected first. Setting CUTLO= and CUTHI= is a compromise between fit and missing data. If you loose too much data, then increase the values. If there is still considerable misfit or skewing of equating, then decrease the values.
Here are the usual effects of CUTLO= and CUTHI=
1. Fit to the Rasch model improves.
2. The count of observations for each person and item decreases.
3. The variance in the data explained by the measures decreases.
CUTLO= is equivalent to the procedure outlined in in Waller, M.I. (1976) "Estimating Parameters in the Rasch Model: Removing the Effects of Random Guessing", Report No. ETS-RB-76-0, Educational Testing Service, Princeton, N.J., https://files.eric.ed.gov/fulltext/ED120261.pdf, also Bruce Choppin. (1983). A two-parameter latent trait model. (CSE Report No. 197). Los Angeles, CA: University of. California, Center for the Study of Evaluation, and David Andrich, Ida Marais, and Stephen Humphry (2012) Using a Theorem by Andersen and the Dichotomous Rasch Model to Assess the Presence of Random Guessing in Multiple Choice Items. Journal of Educational and Behavioral Statistics, 37, 417-442.
Polytomous items: CUTLO= and CUTHI= trim the data relative to the item difficulty, so they tend to remove data in high and low categories. You can adjust the item difficulty relative to the response structure using SAFILE=.
Example 1: Disregard responses where examinees are faced with too great a challenge, and so might guess wildly, i.e., where examinee measure is 2 or more logits lower than item measure:
CUTLO= -2 ; 12% success
This is equivalent to a "Optimum Appropriateness Measurement" (OAM) model in which it is assumed that persons might guess on all the items, so all responses in guessing situations are eliminated.
GUTTMAN SCALOGRAM OF RESPONSES:Table 22
PERSON |ITEM
| 12 22 1 3231311 1322112 2 113322
|62257012473946508491143795368350281
|-----------------------------------
147 +0001111110011010100111111000001000 154
130 +1100101111001110110101000000010100 135
93 +011110010100111010000001101000010 098
129 +111110111011011000000000000001010 134
134 +000110101110010001110010001000010 139
133 +100100111110000000101010011000000 138
137 +100000011110100101101000010000000 143
141 +110100010011100001100000000001010 147
138 +10100101011101000000000010010 144
113 +1011001000101111000000000000010 118
144 +10000010100010001000010001 151
114 +1000010010000100 119
|-----------------------------------
| 12 22 1 3231311 1322112 2 113322
|62257012473946508491143795368350281
Example 2: Richard Gershon applied this technique in Guessing and Measurement with CUTLO=-1 ; 27% success
Example 3: We have some misbehaving children in our sample, but don't want their behavior to distort our final report.
An effective approach is in two stages:
Stage 1. calibrate the items using the good responses
Stage 2. anchor the items and measure the students using all the responses.
In Stage 1, we trim the test. We want to remove the responses by children that are so off-target that successes are probably due to chance or other off-dimensional behavior. These responses will contain most of the misfit. For this we analyze the data using
CUTLO= -2 (choose a suitable value by experimenting)
CUTLO= -1.39 ; 20% success
CUTLO= -1.10 ; 25% success
write an item file from this analysis:
IFILE=if.txt
In Stage 2. Anchor all the items at their good calibrations:
IAFILE=if.txt
Include all the responses (omit CUTLO=)
We can now report all the children without obvious child mis-behavior distorting the item measures.
Help for Winsteps Rasch Measurement and Rasch Analysis Software: www.winsteps.com. Author: John Michael Linacre
Facets Rasch measurement software.
Buy for $149. & site licenses.
Freeware student/evaluation Minifac download Winsteps Rasch measurement software. Buy for $149. & site licenses. Freeware student/evaluation Ministep download |
---|
Forum: | Rasch Measurement Forum to discuss any Rasch-related topic |
---|
Questions, Suggestions? Want to update Winsteps or Facets? Please email Mike Linacre, author of Winsteps mike@winsteps.com |
---|
State-of-the-art : single-user and site licenses : free student/evaluation versions : download immediately : instructional PDFs : user forum : assistance by email : bugs fixed fast : free update eligibility : backwards compatible : money back if not satisfied Rasch, Winsteps, Facets online Tutorials |
---|
Our current URL is www.winsteps.com
Winsteps® is a registered trademark