one of the eas iest to use software : winsteps
DESCRIPTION
One of the eas iest to use Software : Winsteps. www.winsteps.com. 1. 2. 3. 4. Developer of the program : John M. Linacre. Tel/Fax: (312)264-2352. Provides processing of data within Rasch analysis. www.winsteps.com. Introduction. Key options of Winsteps. Test and item analysis - PowerPoint PPT PresentationTRANSCRIPT
LOGO
One of the easiest to use Software: Winsteps
www.winsteps.com
LOGO Introduction
Developer of the program: John M. Linacre1.
Provides processing of data within Rasch analysis
2.
www.winsteps.com3.
Tel/Fax: (312)264-23524.
LOGO Key options of Winsteps
Test and item analysis• Calibration of item difficulty• Investigation of category functioning• Discovering of dimensionality • Construction of scale ability (item map)• Construction of item characteristic curves and item information curves• Differential Item Functioning analysis• Analysis of polytomous response structures (rating scales and partial
credit items)• Fit statistics analysis • etc. Examinee analysis• Analysis of responses of a examinees• etc.
LOGO Work-flow with Winsteps
Control file Data file
Winsteps
GraphsOutput tablesReport Output File
LOGO How to make a control File?
LOGO
The control file tells what analysis you want to do. The template file, TEMPLATE.TXT.
gives you an outline to start from
&INST ; optionalTITLE = "Put your page heading here";Input Data FormatNAME1 = 1 ; column of start of person informationNAMLEN = 30 ; maximum length of person informationITEM1 = ? ; column of first item-level responseNI = ?? ; number of items = test lengthXWIDE = 1 ; number of columns per responsePERSON = Person ; Persons are called ...ITEM = Item ; Items are called ...; DATA = ; data after control specifications
LOGOAn example of a control file for
dichotomous data&INST TITLE = "Biology-1.1" PERSON = Person ; persons are ... ITEM = Item ; items are ... ITEM1 = 2 ; column of response to first item in data record NI = 37 ; number of items NAME1 = 1 ; column of first character of person identifying labelNAMELEN = 21 ; length of person label XWIDE = 1 ; number of columns per item response CODES = 01 ; valid codes in data file UIMEAN = 0 ; item mean for local origin USCALE = 1 ; user scaling for logits UDECIM = 2 ; reported decimal places for user scalingGROUPS=0 ; specify that each item has its own rating scale (partial credit)&END
;Put item labels here for NI= linesA1 A2A3A4A5…
END LABELS
1110111111111101101111111011111110111121100000011110001111110001101111011110…
LOGOAn example of a control file for
polytomous data (PCM)&INST TITLE="PCM" NAME1=1XWIDE=1 ITEM1=11NI=45CODES=012345 GROUPS=0PERSON=PERSONITEM=TASKS&END…
END LABELS
1 0110000001000000001001100100010101000000100012 0000002100210021003002110101200122000010101013 0110010101012114101122100112223101000101221004 2454015433423033535445344443442522324452555255 121001010221233201010210010000311300000022110 …..
LOGOAn example of a control file for
polytomous data (RSM)&INST TITLE="RSM" NAME1=1XWIDE=1 ITEM1=11NI=20CODES=01234NEWSCORE=12345 MODELS=RPERSON=PERSONITEM=TASKS&END…
END LABELS
1 011112012110111221012 242443233233422311233 210104011313011112004 000312001021001302125 11322222211201220 ……………………
LOGO The process is running…
LOGO Getting of outputs
LOGOAn example of an output table (table 3.1
Summary statistics)TABLE 3.1 Русский язык ZOU287WS.TXT Mar 21 10:46 2012INPUT: 1464 Person 34 Item REPORTED: 1464 Person 34 Item 90 CATS WINSTEPS 3.72.3------------------------------------------------------------------------------------ SUMMARY OF 1464 MEASURED Person-------------------------------------------------------------------------------| TOTAL MODEL INFIT OUTFIT || SCORE COUNT MEASURE ERROR MNSQ ZSTD MNSQ ZSTD ||-----------------------------------------------------------------------------|| MEAN 25.0 29.1 .12 .33 1.02 .1 1.01 .0 || S.D. 10.8 5.2 .92 .07 .27 1.0 .33 1.0 || MAX. 55.0 34.0 4.05 1.06 2.28 3.7 4.34 4.7 || MIN. 1.0 7.0 -2.88 .28 .44 -3.0 .34 -2.6 ||-----------------------------------------------------------------------------|| REAL RMSE .36 TRUE SD .84 SEPARATION 2.33 Person RELIABILITY .84 ||MODEL RMSE .34 TRUE SD .85 SEPARATION 2.49 Person RELIABILITY .86 || S.E. OF Person MEAN = .02 |------------------------------------------------------------------------------- VALID RESPONSES: 85.7% (APPROXIMATE)Person RAW SCORE-TO-MEASURE CORRELATION = .94 (approximate due to missing data)CRONBACH ALPHA (KR-20) Person RAW SCORE "TEST" RELIABILITY = .90 (approximate due to missing data) SUMMARY OF 34 MEASURED Item-------------------------------------------------------------------------------| TOTAL MODEL INFIT OUTFIT || SCORE COUNT MEASURE ERROR MNSQ ZSTD MNSQ ZSTD ||-----------------------------------------------------------------------------|| MEAN 1078.3 1255.1 .00 .05 1.00 -.2 1.01 -.1 || S.D. 498.2 167.6 .77 .01 .10 2.7 .15 2.8 || MAX. 2173.0 1442.0 1.56 .07 1.39 9.9 1.67 9.9 || MIN. 416.0 811.0 -1.20 .04 .88 -5.8 .82 -5.5 ||-----------------------------------------------------------------------------|
Estimated person ability Fit statistics
Error of measurement
The number of responses made
Number of correct
responses including extreme scores
Item calibration (difficulty)The average
value of the statistic
Sample standard deviation
Information-weighted fit
statistic
Outlier-sensitive fit statistic
LOGO An example of output table (table 14.1 Item: entry)
TABLE 14.1 Русский язык ZOU287WS.TXT Mar 21 10:46 2012INPUT: 1464 Person 34 Item REPORTED: 1464 Person 34 Item 90 CATS WINSTEPS 3.72.3------------------------------------------------------------------------------------Person: REAL SEP.: 2.33 REL.: .84 ... Item: REAL SEP.: 14.45 REL.: 1.00 Item STATISTICS: ENTRY ORDER -------------------------------------------------------------------------------------------------|ENTRY TOTAL TOTAL MODEL| INFIT | OUTFIT |PT-MEASURE |EXACT MATCH| ||NUMBER SCORE COUNT MEASURE S.E. |MNSQ ZSTD|MNSQ ZSTD|CORR. EXP.| OBS% EXP%| Item G ||------------------------------------+----------+----------+-----------+-----------+------------|| 1 693 1431 .20 .06|1.00 .1|1.03 1.1| .38 .38| 66.0 66.0| R04Q2_1 0 || 2 1075 1431 -1.16 .07| .93 -2.2| .82 -3.6| .44 .34| 77.1 76.9| R04Q2_2 0 || 3 1673 1442 -.25 .04| .99 -.3| .98 -.6| .52 .50| 54.7 51.9| R04Q2_3 0 || 4 850 1425 -.33 .06| .97 -1.4| .94 -1.8| .41 .38| 68.2 67.7| R04Q2_4 0 || 5 822 1371 -.30 .06| .88 -5.8| .83 -5.5| .51 .38| 73.0 67.7| R04Q2_5 0 || 6 1022 1293 .61 .04| .98 -.7| .96 -.9| .53 .51| 51.7 48.8| R04Q2_6 0 || 7 686 1396 .19 .06| .98 -1.2| .97 -1.2| .41 .38| 66.7 65.8| R04Q2_7 0 || 8 2128 1419 -.93 .04|1.01 .2|1.09 1.5| .47 .48| 59.8 60.5| R04Q2_8 0 |
…
| 30 1433 1177 -.60 .05|1.25 6.0|1.26 6.3| .19 .43| 58.2 64.4| R04Q7_1 0 || 31 750 1170 -.49 .07|1.02 .7| .98 -.4| .36 .37| 67.5 69.5| R04Q7_2 0 || 32 624 1092 -.11 .07|1.01 .4|1.00 .0| .37 .38| 66.4 66.8| R04Q7_3 0 || 33 635 1003 1.12 .05|1.20 4.7|1.27 5.1| .35 .48| 49.0 55.2| R04Q7_4 0 || 34 884 811 .85 .04|1.01 .3|1.01 .2| .58 .59| 40.8 40.9| R04Q7_5 0 ||------------------------------------+----------+----------+-----------+-----------+------------|| MEAN 1078.3 1255.1 .00 .05|1.00 -.2|1.01 -.1| | 60.9 60.5| || S.D. 498.2 167.6 .77 .01| .10 2.7| .15 2.8| | 10.6 9.5| |-------------------------------------------------------------------------------------------------
the sum of the correct responses to an item
by the persons
the number of data points used to construct measures
The item difficulty in logits
The standard error for the estimate
Standardized information-
weighted mean square statistic
Standardized outlier-sensitive mean square
statistic
Point-biserial correlation
LOGOAn example of examinee responses (table 7.1
person: responses)
Individual number
Test score
A number of responses with notes of significant deviations(* — significantly negative,+ — significantly positive)
Part А | Part В | Part С
10 1,0111100 11100 11111 11011 11111 11110 10010 10001 10001 12213 ** * * * +
15 2,07 11110 11011 11110 11011 11101 11111 01020 11111 11111 22213 * * * * * * * * +
148 1,21 11111 11111 11111 11111 11111 00000 10021 11002 11111 12000 ***** *
NUMBER - NAME -- POSITION ------ MEASURE - INFIT (MNSQ) OUTFIT 887 11121122211120112112121 1.68 1.4 1.2 RESPONSE: 1: 1 1 1 2 1 1 2 2 2 1Z-RESIDUAL: RESPONSE: 11: 1 1 2 0 1 1 2 1 1 2Z-RESIDUAL: -3 RESPONSE: 21: 1 2 1 2 1 2 2 2 1 1Z-RESIDUAL: -3 RESPONSE: 31: 1 2 1 1…………………………………………Z-RESIDUAL:
Individual number
Significantly negative response
Test score
Fit statistics
LOGO Item map (table 12)TABLE 12.2 Русский язык ZOU000WS.TXT Mar 19 1:19 2012INPUT: 1464 Person 34 Item REPORTED: 1464 Person 34 Item 90 CATS WINSTEPS 3.72.3------------------------------------------------------------------------------------ Person - MAP - Item <more>|<rare> 4 . + | | | | | | 3 . + . | . | . | . | . | . | 2 . T+ .## | .# | .## |T R04Q1_8 R04Q1_9 .### | .#### | .#### | R04Q7_4 1 .####### S+ R04Q1_3 .####### | R04Q7_5 .####### |S R04Q1_5 R04Q8_8 .####### | R04Q1_1 R04Q2_12 R04Q2_6 .########### | R04Q1_2 R04Q1_4 R04Q1_7 R04Q8_4 ########## | .########### M| R04Q2_1 R04Q2_7 0 .########## +M R04Q8_7 .########### | R04Q7_3 .######## | R04Q2_3 R04Q2_4 R04Q2_5 .######## | R04Q7_2 R04Q8_2 ######## | R04Q7_1 R04Q8_3 R04Q8_6 .##### |S R04Q8_5 .##### S| R04Q2_10 R04Q2_8 R04Q8_1 -1 .### + R04Q1_6 R04Q2_9 .### | R04Q2_11 R04Q2_2 .### | .## | .# |T .## T| . | -2 .# + . | . | . | . | | . | -3 + <less>|<frequ> EACH "#" IS 9. EACH "." IS 1 TO 8
Logits
Mean (examinees)
Mean (items)
LOGO Graphs
LOGO Item characteristic curve
Empirical data
Model curve
Confidence interval
LOGO Category Probability Curve
Category “0” Category “1”
Category “2”
LOGO Item Information Function Curve
The biggest error of
measurement
The lowest error of measurement
The biggest error of
measurement
LOGO Test Information Function
The biggest error of
measurement
The biggest error of
measurement
The lowest error of measurement
LOGO
+ Clear graphs and plots +
confidence intervals for model and empirical item characteristic curves (the boundary lines which indicate upper and lower 95% two-sided confidence intervals)
A detailed and easy to use manual
Possibility of examinee responses analysis
- Provides only Rasch
analysis, can not be used for 2Pl or 3Pl analysis
Advantages and limitations of the program
LOGO
www.winsteps.com