Table 10.1 Item statistics in misfit order

(controlled by FITI=, LOCAL=, OUTFIT=, TOTAL=)

 

ITEMS STATISTICS:  MISFIT ORDER

 

 PERSON: REAL SEP.: 1.55  REL.: .70 ... ITEM: REAL SEP.: 3.73  REL.: .93

 

Above the Table are shown the "real" separation coefficient and reliability (separation index) coefficients from Table 3.

 

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

|ENTRY   TOTAL  TOTAL           MODEL|   INFIT  |  OUTFIT  |PTMEASUR-AL|EXACT MATCH|ESTIM| ASYMPTOTE | P-  |     |     |        |                 |

|NUMBER  SCORE  COUNT  MEASURE  S.E. |MNSQ  ZSTD|MNSQ  ZSTD|CORR.  EXP.| OBS%  EXP%|DISCR|LOWER UPPER|VALUE| RMSR|WEIGH|DISPLACE| TAP           G |

|------------------------------------+----------+----------+-----------+-----------+-----+-----------+-----+-----------+--------+-----------------|

|     1     35     35   -6.59A   1.85|1.00   .00|1.00   .04|A .00   .00|100.0 100.0| 1.00|  .00  1.00| 1.00| .000|  .50|    -.52| 1-4           1 |

|     5     31     35   -3.83     .70|1.04   .19| .52   .12|B .55   .55| 88.2  91.7| 1.01|  .05  1.00|  .89| .248| 1.00|    -.87| 2-1-4         2 |

|     6     30     35   -3.38     .64| MAXIMUM MEASURE     |  .53   .58|100.0 100.0|     |           |  .86| .289| 1.00|    -.64| 3-4-1         1 |

|     7     31     35   -3.83     .70| MINIMUM MEASURE     |  .40   .55|100.0 100.0|     |           |  .89| .281| 1.00|     .57| 1-4-3-2       2 |

|     8     27     35                | DROPPED             |           |           |     |           |     |     |     |        | 1-4-2-3       1 |

|    10      0     35                | INESTIMABLE: HIGH   |           |           |     |           |     |     |     |        | 2-4-3-1       3 |

|    11     35     35                | INESTIMABLE: LOW    |           |           |     |           |     |     |     |        | 1-3-1-2-4     4 |

|     9          DELETED             |          |          |           |           |     |           |     |     |     |        | 1-3-2-4       2 |

|    12          DESELECTED          |          |          |           |           |     |           |     |     |     |        | 1-3-2-4-3     1 |

....

|------------------------------------+----------+----------+-----------+-----------+-----+-----------+-----+-----+-----+--------+-----------------|

| MEAN    18.5   35.0    -.59     .94| .96   .04| .68  -.11|           | 89.9  90.0|     |           |     |     |     |     .32|                 |

| P.SD    13.9     .0    4.21     .49| .28   .71| .58   .53|           |  6.3   5.3|     |           |     |     |     |     .27|                 |

--------------------------------------------------------------------------------------------------------------------------------+------------------

 

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

|ENTRY   TOTAL   JMLE   MODEL|   INFIT  |  OUTFIT  |     CMLE      |   CMLE INFIT  |   CMLE OUTFIT |

|NUMBER  SCORE MEASURE  S.E. |MNSQ  ZSTD|MNSQ  ZSTD| MEASURE  S.E. |  MNSQ    ZSTD |  MNSQ    ZSTD |

|----------------------------+----------+----------+---------------+---------------+---------------+

|     5     37   2.42     .22|2.30  5.61|3.62  7.27|   2.32     .22|   2.31    5.65|   3.56    7.26|

|    23     42   2.18     .21|2.41  6.29|4.11  8.97|   2.09     .22|   2.43    6.35|   4.06    8.98|

|    20     50   1.83     .20|1.33  2.00|1.82  3.73|   1.76     .20|   1.34    2.06|   1.81    3.74|

 

Column

Description

ENTRY NUMBER

the sequence number of the person, or item, in your data, and is the reference number used for deletion or anchoring.

TOTAL SCORE

TOTAL COUNT

 

Totalscore=Yes -. This is the score when reading in the data.

 

 

NON-EXTREME SCORE

NON-EXTREME COUNT

Totalscore=No -  the raw score and count of response by a person on the test, or the sum of the scored responses to an item by the persons, omitting responses in extreme and inestimable scores. This is the score when estimating person abilities and item difficulties. Scored responses are transformed (re-counted) so that the lowest response is zero.

MEASURE

the estimate (or calibration) of the person ability (theta, B, beta, etc.), or the item difficulty (b, D, delta, etc.). Values are reported in logits with two decimal places, unless rescaled by USCALE=, UIMEAN=, UPMEAN=, UDECIM=.

 

The difficulty of an item is defined to be the point on the latent variable at which its high and low categories are equally probable. SAFILE= can be used to alter this definition.

 
If unexpected results are reported, check whether TARGET= or CUTLO= or CUTHI= or ISGROUPS= are specified.

A after MEASURE, MAXIMUM, etc.

see STATUS Table below

MODEL S.E.

REAL S.E.

MODEL S.E. is the standard error of the estimate.  REAL S.E. is the misfit-inflated standard error. These are commonly referred to as conditional standard errors of measurement (CSEM).

Real S.E>  while you are improving your results. This assumes misfit contradicts the Rasch model. Model S.E. when your results are as good as they can be. This assumes misfit is the randomness predicted by the Rasch model.

INFIT

an information-weighted statistic, which is more sensitive to unexpected behavior affecting responses to items near the person's measure level.

OUTFIT

an unweighted statistic, more sensitive to unexpected behavior by persons on items far from the person's measure level.

MNSQ

a mean-square statistic computed for all scores responses, excluding responses in extreme total scores. This is a chi-square statistic divided by its degrees of freedom. Its expectation is 1.0.  Values substantially less than 1.0 indicate overfit = dependency in your data. Values values substantially greater than 1.0 indicate underfit = unmodeled noise.  See dichotomous and polytomous fit statistics.

 

Value

Meaning

>2.0

Off-variable noise is greater than useful information.  Degrades measurement. Always remedy the large misfits first.

>1.5

Noticeable off-variable noise.  Neither constructs nor degrades measurement

0.5 - 1.5

Productive of measurement

<0.5

Overly predictable. Misleads us into thinking we are measuring better than we really are. (Attenuation paradox.). Item misfits <1.0 are only of concern when shortening a test. Person misfits <1.0 are rarely of concern.

ZSTD

ZEMP

LOG

PROB

the INFIT or OUTFIT mean-square fit statistic t standardized to approximate a theoretical "unit normal", mean 0 and variance 1, distribution. ZSTD (standardized as a z-score) is used of a t-test result when either the t-test value has effectively infinite degrees of freedom (i.e., approximates a unit normal value) or the Student's t-statistic distribution value has been adjusted to a unit normal value. The standardization is shown on RSA, p.100-101. When LOCAL=Y, then ZEMP is shown, indicating a local {0,1} standardization.  When LOCAL=LOG, then LOG is shown, and the natural logarithms of the mean-squares are reported. More exact values are shown in the Output Files. When LOCAL=PROB, the probability of the mean-square is shown.
Ben Wright advises: "ZSTD is only useful to salvage non-significant MNSQ>1.5, when sample size is small or test length is short."

PTMEASUR-AL CORR.

PTMEASUR-AL EXP.

an observed point-correlation: PTBISERL-AL, PTBISERL-EX, PTMEASURE-A, PTMEASUR-EX, see Correlations.  Negative reported correlations suggest that the orientation of the scoring on the item, or by the person, may be opposite to the orientation of the latent variable. This may be caused by item miskeying, reverse scoring, person special knowledge, guessing, data entry errors, or the expected randomness in the data.

 

Correlations of 0.00 may be because the correlation cannot be calculated due to the structure of the data.

 

In FIT ORDER Tables 6.1 and 10.1, letters A, B, ... indicating the identity of persons or items appearing on the Infit and Outfit plots, Tables 4, 5, 8, 9,  precede the correlations. If the point-correlation is inestimable because items have different numbers of categories, this column does not appear, or only the FIT letters display.

 

EXP. is the expected value of the point-correlation when the data fit the Rasch model with the estimated measures. See Correlations.

EXACT MATCH OBS%

EXACT MATCH EXP%  

OBServed% is the percent of data points that are within 0.5 score points of their expected values, i.e., that match predictions.

EXPected% is the percent of data points that are predicted to be within 0.5 score points of their expected values.

ESTIM DISCRIM

an estimate of the 2-PL item discrimination, see DISCRIM=

Negative discriminations are usually problematic and accompanied by negative point-biserial correlations. These indicate that the scoring on the item may be contradicting the overall latent variable. However, this is not a universal rule, so please look at the infit and outfit mean-squares to see whether they also indicate problems (values much greater than 1.0). Exceptions to the rule include very hard and very easy items, and situations where the person sample variance is very small.

ASYMPTOTE LOWER

ASYMPTOTE UPPER

estimates of the upper and lower asymptotes for dichotomous items, see ASYMPTOTE=

P-VALUE

the observed proportion-correct (p-value) for 0/1 dichotomies, or observed average rating on the item, see PVALUE=.

RMSR

root-mean-square-residual of observations not in extreme scores, see RMSR=.

WMLE

Warm's Mean Likelihood Estimate, see WMLE=

QCMLE

Quasi-Conditional Maximum Likelihood Estimate, see QCMLE=

WEIGH

the weight assigned by IWEIGHT= or PWEIGHT=. When WEIGHT = 0.0, the item or person is estimated, but does not influence the estimate of any other item or person.

CMLE MEASURE

CMLE item measure or CMLE-based AMLE person measure when CMLE=Yes

CMLE S.E.

CMLE measure S.E. for items and person when CMLE=Yes

CMLE INFIT MNSQ

CMLE Infit Mean-square fit computed from CMLE probabilities when CMLE=Yes

CMLE INFIT ZSTD

CMLE Infit Z-standardized fit computed from CMLE probabilities when CMLE=Yes

CMLE OUTFIT MNSQ

CMLE Outfitfit Mean-square fit computed from CMLE probabilities when CMLE=Yes

CMLE OUTFIT ZSTD

CMLE Outfit Z-standardized fit computed from CMLE probabilities when CMLE=Yes

CMLE WMLE

CMLE Warm's Weighted Mean Likelihood estimates when CMLE=Yes and WMLE=Yes

DISPLACE

the displacement of the reported MEASURE from its data-derived value. This should only be shown with anchored measures. The displacement values can be see in IFILE= and PFILE= output files.

The displacement is an estimate of the amount to add to the MEASURE to make it conform with the data.
+ logit displacement for a person ability indicates that the observed person score is higher (higher ability) than the expected person score based on the reported measure (anchor value).
+ logit displacement for an item difficulty indicates that the observed item score is lower (harder item) than the expected item score based on the reported measure (anchor value).
Unanchored measures: If small displacements are shown, tighten the convergence criteria, CONVERGE=, LCONV=, RCONV=
Anchored measures: we expect half the displacements to be negative and half to be positive, and for them to be normally distributed according to the standard errors of the measures.

PERSON

ITEM

the name of the list of persons (data rows) or items (data columns) reported here

G

the grouping code assigned with ISGROUPS=. Table 3.2, 3.3, etc. show the details of rating scales, etc., for each grouping code.

M

model code assigned with MODELS=

SUBSET:

the person or items are in incomparable subsets, see Subsets.

 

Reported

STATUS

PFILE=, IFILE=

Measured?

Reason

A (after Measure)

2

Yes

Anchored (fixed) measure. The reported S.E. is that which would have been obtained if the value had been estimated. Values are reported in logits with two decimal places, unless rescaled by USCALE=, UDECIM=

(MEASURE)

1

Yes

Estimated measure

MINIMUM

0

Yes

Extreme minimum score. Measure estimated using EXTRSC=

MINIMUM and MAXIMUM measures may occur because other MINIMUM and MAXIMUM measures have been dropped from the estimation. Inspect with TOTALSCORE=No.

MAXIMUM

-1

Yes

Extreme maximum score. Measure estimated using EXTRSC=

MINIMUM and MAXIMUM measures may occur because other MINIMUM and MAXIMUM measures have been dropped from the estimation.  Inspect with TOTALSCORE=No.

DROPPED

-2

No

No responses available for measurement. Perhaps due to CUTLO=, CUTHI=, CODES=, or deletion of other persons or items.

Also check that anchored items. IAFILE=, has matching SAFILE= for all items.

DELETED

-3

No

Deleted by user. PDELETE=, PDFILE=, IDELETE=, IDFILE=, PSELECT=, ISELECT=

INESTIMABLE: HIGH

-4

No

For an item, this can be resolved using SAFILE= or grouping this item with a similar estimable item:

SAFILE=*

23 0 0 ; item-number bottom-category 0

23 1 0 ; item-number top-category 0

*

For a person: change the observation by this person on the hardest item into a lower category.

INESTIMABLE: LOW

-5

No

The measure probably has a low value. For an item, this can be resolved using SAFILE=  or grouping this item with a similar estimable item:

SAFILE=*

23 0 0 ; item-number bottom-category 0

23 1 0 ; item-number top-category 0

*

For a person: change the observation by this person on the easiest item into a higher category.

DROPPED

A (after Measure)

-6

Yes

Anchored (fixed) measure with no observed raw score

DESELECTED

-7 to -16

No

Temporarily deselected by Specification box with iSELECT= , etc. (usual STATUS - 10)

OMITTED

-17 to -26

No

Temporarily deleted by Specification box with iDELETE=, etc. (usual STATUS - 20)

REMOVED

-27 to -36

No

Temporarily deselected and deleted by Specification box with iSELECT= , etc., and iDELETE=, etc. (usual STATUS - 30)

 

Example: To eliminate misfitting items, output Table 10.1. Copy into Notepad++ (freeware) or other software capable of rectangular copy (alt+mouse). Then paste into your Winsteps control file

IDFILE=*

pasted entry numbers - put enough blank lines here first for all the numbers

*

 

Another approach is to screen out very unexpected answers, rather than whole people. Do this with CUTLO= -2


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

Rasch Books and Publications: Winsteps and Facets
Applying the Rasch Model (Winsteps, Facets) 4th Ed., Bond, Yan, Heene Advances in Rasch Analyses in the Human Sciences (Winsteps, Facets) 1st Ed., Boone, Staver Advances in Applications of Rasch Measurement in Science Education, X. Liu & W. J. Boone Rasch Analysis in the Human Sciences (Winsteps) Boone, Staver, Yale Appliquer le modèle de Rasch: Défis et pistes de solution (Winsteps) E. Dionne, S. Béland
Introduction to Many-Facet Rasch Measurement (Facets), Thomas Eckes Rasch Models for Solving Measurement Problems (Facets), George Engelhard, Jr. & Jue Wang Statistical Analyses for Language Testers (Facets), Rita Green Invariant Measurement with Raters and Rating Scales: Rasch Models for Rater-Mediated Assessments (Facets), George Engelhard, Jr. & Stefanie Wind Aplicação do Modelo de Rasch (Português), de Bond, Trevor G., Fox, Christine M
Exploring Rating Scale Functioning for Survey Research (R, Facets), Stefanie Wind Rasch Measurement: Applications, Khine Winsteps Tutorials - free
Facets Tutorials - free
Many-Facet Rasch Measurement (Facets) - free, J.M. Linacre Fairness, Justice and Language Assessment (Winsteps, Facets), McNamara, Knoch, Fan
Other Rasch-Related Resources: Rasch Measurement YouTube Channel
Rasch Measurement Transactions & Rasch Measurement research papers - free An Introduction to the Rasch Model with Examples in R (eRm, etc.), Debelak, Strobl, Zeigenfuse Rasch Measurement Theory Analysis in R, Wind, Hua Applying the Rasch Model in Social Sciences Using R, Lamprianou Journal of Applied Measurement
Rasch Models: Foundations, Recent Developments, and Applications, Fischer & Molenaar Probabilistic Models for Some Intelligence and Attainment Tests, Georg Rasch Rasch Models for Measurement, David Andrich Constructing Measures, Mark Wilson Best Test Design - free, Wright & Stone
Rating Scale Analysis - free, Wright & Masters
Virtual Standard Setting: Setting Cut Scores, Charalambos Kollias Diseño de Mejores Pruebas - free, Spanish Best Test Design A Course in Rasch Measurement Theory, Andrich, Marais Rasch Models in Health, Christensen, Kreiner, Mesba Multivariate and Mixture Distribution Rasch Models, von Davier, Carstensen
As an Amazon Associate I earn from qualifying purchases. This does not change what you pay.

facebook Forum: Rasch Measurement Forum to discuss any Rasch-related topic

To receive News Emails about Winsteps and Facets by subscribing to the Winsteps.com email list,
enter your email address here:

I want to Subscribe: & click below
I want to Unsubscribe: & click below

Please set your SPAM filter to accept emails from Winsteps.com
The Winsteps.com email list is only used to email information about Winsteps, Facets and associated Rasch Measurement activities. Your email address is not shared with third-parties. Every email sent from the list includes the option to unsubscribe.

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


 

 
Coming Rasch-related Events
May 17 - June 21, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
June 12 - 14, 2024, Wed.-Fri. 1st Scandinavian Applied Measurement Conference, Kristianstad University, Kristianstad, Sweden http://www.hkr.se/samc2024
June 21 - July 19, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Further Topics (E. Smith, Winsteps), www.statistics.com
Aug. 5 - Aug. 6, 2024, Fri.-Fri. 2024 Inaugural Conference of the Society for the Study of Measurement (Berkeley, CA), Call for Proposals
Aug. 9 - Sept. 6, 2024, Fri.-Fri. On-line workshop: Many-Facet Rasch Measurement (E. Smith, Facets), www.statistics.com
Oct. 4 - Nov. 8, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
Jan. 17 - Feb. 21, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
May 16 - June 20, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
June 20 - July 18, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Further Topics (E. Smith, Facets), www.statistics.com
Oct. 3 - Nov. 7, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com

 

 

Our current URL is www.winsteps.com

Winsteps® is a registered trademark