Search results for: asymmetric information.
2 Comparing Test Equating by Item Response Theory and Raw Score Methods with Small Sample Sizes on a Study of the ARTé: Mecenas Learning Game
Authors: Steven W. Carruthers
Abstract:
The purpose of the present research is to equate two test forms as part of a study to evaluate the educational effectiveness of the ARTé: Mecenas art history learning game. The researcher applied Item Response Theory (IRT) procedures to calculate item, test, and mean-sigma equating parameters. With the sample size n=134, test parameters indicated “good” model fit but low Test Information Functions and more acute than expected equating parameters. Therefore, the researcher applied equipercentile equating and linear equating to raw scores and compared the equated form parameters and effect sizes from each method. Item scaling in IRT enables the researcher to select a subset of well-discriminating items. The mean-sigma step produces a mean-slope adjustment from the anchor items, which was used to scale the score on the new form (Form R) to the reference form (Form Q) scale. In equipercentile equating, scores are adjusted to align the proportion of scores in each quintile segment. Linear equating produces a mean-slope adjustment, which was applied to all core items on the new form. The study followed a quasi-experimental design with purposeful sampling of students enrolled in a college level art history course (n=134) and counterbalancing design to distribute both forms on the pre- and posttests. The Experimental Group (n=82) was asked to play ARTé: Mecenas online and complete Level 4 of the game within a two-week period; 37 participants completed Level 4. Over the same period, the Control Group (n=52) did not play the game. The researcher examined between group differences from post-test scores on test Form Q and Form R by full-factorial Two-Way ANOVA. The raw score analysis indicated a 1.29% direct effect of form, which was statistically non-significant but may be practically significant. The researcher repeated the between group differences analysis with all three equating methods. For the IRT mean-sigma adjusted scores, form had a direct effect of 8.39%. Mean-sigma equating with a small sample may have resulted in inaccurate equating parameters. Equipercentile equating aligned test means and standard deviations, but resultant skewness and kurtosis worsened compared to raw score parameters. Form had a 3.18% direct effect. Linear equating produced the lowest Form effect, approaching 0%. Using linearly equated scores, the researcher conducted an ANCOVA to examine the effect size in terms of prior knowledge. The between group effect size for the Control Group versus Experimental Group participants who completed the game was 14.39% with a 4.77% effect size attributed to pre-test score. Playing and completing the game increased art history knowledge, and individuals with low prior knowledge tended to gain more from pre- to post test. Ultimately, researchers should approach test equating based on their theoretical stance on Classical Test Theory and IRT and the respective assumptions. Regardless of the approach or method, test equating requires a representative sample of sufficient size. With small sample sizes, the application of a range of equating approaches can expose item and test features for review, inform interpretation, and identify paths for improving instruments for future study.Keywords: Effectiveness, equipercentile equating, IRT, learning games, linear equating, mean-sigma equating.
Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 10061 Using Statistical Significance and Prediction to Test Long/Short Term Public Services and Patients Cohorts: A Case Study in Scotland
Authors: Sotirios Raptis
Abstract:
Health and Social care (HSc) services planning and scheduling are facing unprecedented challenges, due to the pandemic pressure and also suffer from unplanned spending that is negatively impacted by the global financial crisis. Data-driven approaches can help to improve policies, plan and design services provision schedules using algorithms that assist healthcare managers to face unexpected demands using fewer resources. The paper discusses services packing using statistical significance tests and machine learning (ML) to evaluate demands similarity and coupling. This is achieved by predicting the range of the demand (class) using ML methods such as Classification and Regression Trees (CART), Random Forests (RF), and Logistic Regression (LGR). The significance tests Chi-Squared and Student’s test are used on data over a 39 years span for which data exist for services delivered in Scotland. The demands are associated using probabilities and are parts of statistical hypotheses. These hypotheses, as their NULL part, assume that the target demand is statistically dependent on other services’ demands. This linking is checked using the data. In addition, ML methods are used to linearly predict the above target demands from the statistically found associations and extend the linear dependence of the target’s demand to independent demands forming, thus, groups of services. Statistical tests confirmed ML coupling and made the prediction statistically meaningful and proved that a target service can be matched reliably to other services while ML showed that such marked relationships can also be linear ones. Zero padding was used for missing years records and illustrated better such relationships both for limited years and for the entire span offering long-term data visualizations while limited years periods explained how well patients numbers can be related in short periods of time or that they can change over time as opposed to behaviours across more years. The prediction performance of the associations were measured using metrics such as Receiver Operating Characteristic (ROC), Area Under Curve (AUC) and Accuracy (ACC) as well as the statistical tests Chi-Squared and Student. Co-plots and comparison tables for the RF, CART, and LGR methods as well as the p-value from tests and Information Exchange (IE/MIE) measures are provided showing the relative performance of ML methods and of the statistical tests as well as the behaviour using different learning ratios. The impact of k-neighbours classification (k-NN), Cross-Correlation (CC) and C-Means (CM) first groupings was also studied over limited years and for the entire span. It was found that CART was generally behind RF and LGR but in some interesting cases, LGR reached an AUC = 0 falling below CART, while the ACC was as high as 0.912 showing that ML methods can be confused by zero-padding or by data’s irregularities or by the outliers. On average, 3 linear predictors were sufficient, LGR was found competing well RF and CART followed with the same performance at higher learning ratios. Services were packed only when a significance level (p-value) of their association coefficient was more than 0.05. Social factors relationships were observed between home care services and treatment of old people, low birth weights, alcoholism, drug abuse, and emergency admissions. The work found that different HSc services can be well packed as plans of limited duration, across various services sectors, learning configurations, as confirmed by using statistical hypotheses.
Keywords: Class, cohorts, data frames, grouping, prediction, probabilities, services.
Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 435