Working with Atypical Samples

Cui, Zhongmin

doi:10.1111/emip.12360

Cited by 2 publications

(2 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to both applications, most data we have in the field of educational measurement seem to be small, with a possible exception of log data or process data that are generated when a test taker or learner interacts with a computer. Worse yet, we may even see smaller data than previously due to various reasons including COVID‐19 (Babcock & Hodge, 2020; Cui, 2020; König et al., 2021; Livingston & Kim, 2009; Puhan et al., 2007). One might ask: can we still make the most of machine learning with small data?…”

Section: Big Data and Small Datamentioning

confidence: 71%

Machine Learning and Small Data

Cui

2021

Educational Measurement

Self Cite

View full text Add to dashboard Cite

Commonly used machine learning applications seem to relate to big data. This article provides a gentle review of machine learning and shows why machine learning can be applied to small data too. An example of applying machine learning to screen irregularity reports is presented. In the example, the support vector machine and multinomial naïve Bayes methods were used and compared. The performance of machine learning was compared to human experts in terms of flagging records to be excluded from equating. The application of machine learning seemed to be successful, although the data only consisted of a couple of thousand records. Recommendations in using machine learning are provided.

show abstract

Section: Big Data and Small Datamentioning

confidence: 71%

Machine Learning and Small Data

Cui

2021

Educational Measurement

Self Cite

View full text Add to dashboard Cite

show abstract

“…For example, it caused large‐scale extended school closures and disrupted student learning and assessment to an unprecedented extent. Starting from the early months (spring of 2020 and onward), assessment professionals expressed concerns and offered suggestions for equating during the COVID‐19 pandemic (e.g., Cui, 2020; Keng et al., 2020; NCME, 2021).…”

mentioning

confidence: 99%

Evaluating Population Invariance of Test Equating During the COVID‐19 Pandemic

Li¹,

Kapoor²

2022

Educational Measurement

View full text Add to dashboard Cite

Population invariance is a desirable property of test equating which might not hold when significant changes occur in the test population, such as those brought about by the COVID-19 pandemic. This research aims to investigate whether equating functions are reasonably invariant when the test population is impacted by the pandemic. Based on pseudo-test forms constructed from an operational form administered in the springs of 2019, 2020, and 2021, this study conducted preequating and postequating using different data collection designs and different sample sizes based on each year's data. Raw-to-scale score conversion tables from each equating and the group means after applying these conversions were compared with those from a criterion equating, that is, single group design postequating based on the 2019 data. Within each design and sample size condition, the magnitude of the differences between the 2021 equating and the criterion equating was mostly similar to the magnitude of differences between the 2020 equating and the criterion, indicating a reasonable extent of invariance in equating results. Nevertheless, some equating designs showed slightly less invariance than others.

show abstract

Working with Atypical Samples

Cited by 2 publications

References 11 publications

Machine Learning and Small Data

Machine Learning and Small Data

Evaluating Population Invariance of Test Equating During the COVID‐19 Pandemic

Contact Info

Product

Resources

About