Bio

Mirian Lima is currently collaborating as an independent consultant with two higher studies institutes in Mindelo, Cape Verde, in order to introduce computational thinking in their curriculum. She received an MS in social research methods and economic development analysis from the London School of Economics, and has worked in public policy design and evaluation for over a decade. Her main areas of interest are education policies to counter the twenty-first-century knowledge gap in developing nations, open data initiatives, and data science for social good and its applications to economic development, in particular.

Computational Essay

Ideas & Growth »

Project: Framework of Adaptive Testing in Mathematica

Goal of the project:

Design a demo of a basic framework of a computer-adaptive test in Mathematica.

Summary of work:

Creation of a dynamic interface design to answer questions using FormFunction and CloudExpressions, which interact with a test dataset in order to produce data in a small matrix of persons (rows) by questions (items/columns).

The graph image shown is the simplest IRT model for a dichotomous item with only one item parameter, known as the one-parameter logistic function. Then through an iterative model (the Rasch algorithm), the item response function is continuously adjusted to the squared sum of residuals for the entire matrix. Once the squared sum of residuals of the entire matrix is close to zero, the student ability and item difficulty estimate are used as parameters in a question selection algorithm that identifies the next question with the maximum likelihood of success.

Results and future work:

Work to be done includes developing:

1. an initial form that registers new student IDs and a final form step that exits the test if a certain threshold has been met (either a certain number of questions or a certain level of ability);
2. an algorithm that updates the student ability value for each test taker as well as the item difficulty values for each question (item) in the test;
3. a third block of code that should use the updated student ability and item difficulty parameters to estimate likelihoods for the remaining questions, and select the one with the highest likelihood as the next one.