Typical Use of eatATA: a Minimal Example

Benjamin Becker

2023-12-12

eatATA efficiently translates test design requirements for Automated Test Assembly (ATA) into constraints for a Mixed Integer Linear Programming Model (MILP). A number of efficient and user-friendly functions are available that translate conceptual test assembly constraints to constraint objects for MILP solvers, like the GLPK solver. In the remainder of this vignette we will illustrate the use of eatATA using a minimal example. A general overview over eatATA can be found in the vignette Overview of eatATA Functionality.

Setup

The eatATA package can be installed from CRAN.

install.packages("eatATA")

Item Pool

First, eatATA is loaded into your R session. In this vignette we use a small simulated item pool, items_mini. The goal will be to assemble a single test form consisting of ten items, an average test time of eight minutes and maximum TIF at medium ability. We therefore calculate the IIF at medium ability and append it to the item pool using the calculateIFF() function.

# loading eatATA
library(eatATA)

# item pool structure
str(items_mini)
#> 'data.frame':    30 obs. of  4 variables:
#>  $ item      : int  1 2 3 4 5 6 7 8 9 10 ...
#>  $ format    : chr  "mc" "mc" "mc" "mc" ...
#>  $ time      : num  27.8 15.5 31 29.9 23.1 ...
#>  $ difficulty: num  -1.881 0.843 1.119 0.729 -0.489 ...

# calculate and append IIF
items_mini[, "IIF_0"] <- calculateIIF(B = items_mini$difficulty, theta = 0)

In Table 1 you can see the first five items of the item pool.

Table 1. First 5 Items of the Item Pool
item format time difficulty IIF_0
1 mc 27.786 -1.881 0.1090032
2 mc 15.453 0.843 0.4494582
3 mc 31.016 1.119 0.3266106
4 mc 29.874 0.729 0.5033924
5 mc 23.134 -0.489 0.6108816

Objective Function

Next, the objective function is defined: The TIF should be maximized at medium ability. For this, we use the maxObjective() function.

testInfo <- maxObjective(nForms = 1, itemValues = items_mini$IIF,
                          itemIDs = items_mini$item)

Constraints

Our further, fixed constraints are defined as additional constraint objects.

itemNumber <- itemsPerFormConstraint(nForms = 1, operator = "=", 
                                     targetValue = 10, 
                                     itemIDs = items_mini$item)

itemUsage <- itemUsageConstraint(nForms = 1, operator = "<=", 
                                 targetValue = 1, 
                                 itemIDs = items_mini$item)

testTime <- itemValuesDeviationConstraint(nForms = 1, 
                                itemValues = items_mini$time,
                                targetValue = 8 * 60, 
                                allowedDeviation = 5,
                                relative = FALSE, 
                                itemIDs = items_mini$item)

Alternatively, we could determine the appropriate test time based on the item pool using the autoItemValuesMinMax() function.

testTime2 <- autoItemValuesMinMaxConstraint(nForms = 1, 
                                itemValues = items_mini$time,
                                testLength = 10, 
                                allowedDeviation = 5,
                                relative = FALSE, 
                                itemIDs = items_mini$item)
#> The minimum and maximum values per test form are: min = 418.09 - max = 428.09

Solver usage

To automatically assemble the test form based on our constraints, we call the useSolver() function. In this function we define which solver should be used as back end. As a default solver, we recommend GLPK, which is automatically installed alongside this package.

solver_out <- useSolver(list(itemNumber, itemUsage, testTime, testInfo),
                        solver = "GLPK")
#> GLPK Simplex Optimizer 5.0
#> 34 rows, 31 columns, 151 non-zeros
#>       0: obj =  -0.000000000e+00 inf =   4.850e+02 (2)
#>      14: obj =  -0.000000000e+00 inf =   0.000e+00 (0)
#> *    34: obj =   6.734471402e+00 inf =   4.441e-16 (0)
#> OPTIMAL LP SOLUTION FOUND
#> GLPK Integer Optimizer 5.0
#> 34 rows, 31 columns, 151 non-zeros
#> 30 integer variables, all of which are binary
#> Integer optimization begins...
#> Long-step dual simplex will be used
#> +    34: mip =     not found yet <=              +inf        (1; 0)
#> +    44: >>>>>   6.579408205e+00 <=   6.732773863e+00   2.3% (9; 0)
#> +    46: >>>>>   6.729573876e+00 <=   6.729573876e+00   0.0% (7; 5)
#> +    46: mip =   6.729573876e+00 <=     tree is empty   0.0% (0; 19)
#> INTEGER OPTIMAL SOLUTION FOUND

Inspect solution

The solution can be inspected directly via inspectSolution() or appended to the item pool via appendSolution(). Using the inspectSolution() function an additional row is created that calculates the column sums for all numeric variables.

inspectSolution(solver_out, items = items_mini, idCol = "item")
#> $form_1
#>     item format      time  difficulty   theta=0
#> 8      8     mc  30.21856 -0.36707654 0.6564876
#> 14    14   open  62.99738  0.58136415 0.5712686
#> 15    15   open  56.59458 -0.12012428 0.7150196
#> 20    20   open  87.05063  0.10201223 0.7170949
#> 22    22  order  39.92415  0.15006395 0.7108712
#> 24    24  order  40.52289 -0.53606969 0.5910511
#> 25    25  order  52.15832  0.14083641 0.7122442
#> 26    26  order  38.29060  0.02381911 0.7222039
#> 28    28  order  43.77592  0.41298287 0.6403034
#> 29    29  order  25.55363  0.24091747 0.6930294
#> Sum  211   <NA> 477.08666  0.62872568 6.7295739
appendSolution(solver_out, items = items_mini, idCol = "item")
#>    item format     time  difficulty    theta=0 form_1
#> 1     1     mc 27.78586 -1.88090278 0.10900318      0
#> 2     2     mc 15.45258  0.84295865 0.44945822      0
#> 3     3     mc 31.01590  1.11881538 0.32661056      0
#> 4     4     mc 29.87421  0.72867743 0.50339241      0
#> 5     5     mc 23.13401 -0.48870993 0.61088162      0
#> 6     6     mc 25.19305  0.47273874 0.61733915      0
#> 7     7     mc 25.66340 -1.18054268 0.30183441      0
#> 8     8     mc 30.21856 -0.36707654 0.65648760      1
#> 9     9     mc 26.61642 -0.56879434 0.57682871      0
#> 10   10     mc 15.35510  1.35397237 0.23900562      0
#> 11   11   open 65.85163 -0.75879786 0.48917461      0
#> 12   12   open 35.94400  2.49927381 0.04012039      0
#> 13   13   open 78.85030  1.33165799 0.24650909      0
#> 14   14   open 62.99738  0.58136415 0.57126860      1
#> 15   15   open 56.59458 -0.12012428 0.71501958      1
#> 16   16   open 45.12778 -1.28629686 0.26229560      0
#> 17   17   open 48.11908 -0.86124314 0.44088544      0
#> 18   18   open 76.32293  0.76977036 0.48398822      0
#> 19   19   open 76.20244 -1.39388826 0.22601541      0
#> 20   20   open 87.05063  0.10201223 0.71709486      1
#> 21   21  order 22.47400 -0.43147145 0.63341304      0
#> 22   22  order 39.92415  0.15006395 0.71087118      1
#> 23   23  order 57.71593 -0.82071059 0.45992776      0
#> 24   24  order 40.52289 -0.53606969 0.59105111      1
#> 25   25  order 52.15832  0.14083641 0.71224418      1
#> 26   26  order 38.29060  0.02381911 0.72220392      1
#> 27   27  order 45.97548  2.79595336 0.02450104      0
#> 28   28  order 43.77592  0.41298287 0.64030341      1
#> 29   29  order 25.55363  0.24091747 0.69302944      1
#> 30   30  order 19.50162 -0.51434114 0.60026891      0