Tutorial 4: CTT and Fit Statistics(2013731) 

Home 1. Install R 2. Install TAM 3. Rasch Model 4. CTT, Fit 5. Partial Credit Model 6. Population Model 7. DIF More resources 
Summary of TutorialThis tutorial shows how to score raw item responses, compute CTT (classical test theory) statistics, compute fit statistics and plot expected scores curves. The R script file for this tutorial can be downloaded through the link Tutorial4.R . Data filesThe data file is the same as for Tutorial 3, except that for multiplechoice items, raw responses are captured rather than scored responses. For example, if a student chose option 2 for Question 3, then 2 is recorded even though option 4 is the correct answer. The test paper and the data file can be downloaded through the following links: Numeracy test paperNumeracy data file in csv format To run an IRT analysis using TAM, the data need to be scored first. R script fileLoad R librariesBefore TAM functions can be called, the TAM library has to be loaded into the R workspace. Line 2 loads the TAM library. Read data file
Line 5 sets the working directory and line 6 reads the data file "D1_resp.csv"
into a variable called "raw_resp". Score data file
Line 9 specifies the "keys" of the item responses.
Line 10 scores an item response to "1" if the item response matches the key, otherwise scores it "0". Run IRT analysis (MML)Line 13 runs an IRT analysis using MML estimation. The results of the IRT analysis
are stored in variable "mod1". Compute ability estimatesLine 16 computes the weighted likelihood ability estimates (WLE). The ability estimates and
respondent test scores are stored in variable "Abil". CTT
After the IRT analysis is run, the ctt function can be called. Line 19 calls the tam.ctt function.
The result of the function is stored in variable "ctt1". In the Console window, you can check the ctt results by typing "ctt1" followed by ENTER. Below is an excerpt of the D1_ctt1.csv file. The columns of the D1_ctt1.csv file show item name, number of respondents, item category, response frequency, percentage, pointbiserial correlation using WLE, the mean and standard deviation of WLE abilities for respondents in the response category. As an exercise, you can run the ctt function for the scored data. For example, ctt2 < tam.ctt(scored, Abil$theta) IRT person separation reliability
IRT person separation reliability is stored in the WLE.rel variable (see
pdf document on tam.wle). To
retrieve it, type Iitem fit statistics
Item fit statistics can be obtained after IRT analysis is run (see pdf document on tam.fit).
The residualbased item fit statistics are produced. Line 23 of the R script file
computes the item fit statistics. Plot Expected Scores Curves
This section shows how expected scores curves can be plotted in R. In the case of
the dichotomous items, the expected scores curves are the ICC (item
characterisitc curves) ExercisesThe following is a data set from FIMS (First International Mathematics Study, IEA), Test A (partial), for populations 1a and 1b, for Australia. Carry out an IRT analysis using this data set.Note that the first column of the data is the "gender" variable. The item responses are in columns 215. Openended questions have been scored (0/1). Multiplechoice items have not been scored and raw responses are recorded (A=1, B=2, etc.). FIMS Test A (part) Questions FIMS Test A Australian data 