Identifying At-Risk Students

 Apthorp, H. A., Dean, C. S., Florian, J. E., Lauer, P. A., Reichardt, R., Sanders, N. M., & Snow-Renner,R. (2001). Helping At-Risk Students Meet Standards.  A Synthesis of Evidence-Based Classroom Practices. Aurora, CO: Mid-continent Research for Education and Learning.

Introduction

Effective teachers in the school classroom context continually monitor their students using informal tests, discussions and observations to detect students at risk of failing their subject. They are then in a position to assist the student, perhaps directing them to additional resources, encouraging them or even seeking assistance from learning support specialists. Teachers in university settings, where classes may be in the order of several hundred students are often unable to establish the type of student/teacher relationship that

exists in a school setting. Identifying those students in the class who lack confidence, motivation or knowledge becomes difficult and is exacerbated when students study in a distance mode. Rather than use observation of students, teachers in this context need to rely almost entirely on student responses to pre-course measures and perhaps early assessment items. In many courses, students who perform poorly in pre-course measures (for example pre-tests) are targeted for intervention programs.

Purpose

The current study focuses on the performance of students within a Tertiary

Preparation Program (TPP). This program has been designed for students returning to study after an absence and is often composed of students who are mature aged (median 30 years) and who have not had success in studying mathematics. In this study we employ an existing mathematics pre-test (called the M-test) to assess student’s prior knowledge in mathematics along with measures of student’s self-efficacy and some demographic variables. The latter measure is included as it is known to predict performance in mathematics for older and lower achieving students

Population/Sample

 One hundred and twenty five students participated in the study out of a total of 300 enrolled students. Of these respondents, 47 were male and 78 were female, with ages ranging from 18 to 54 years (median: 30 years). The number of years since the student last studied mathematics formally ranged from 0 to 40 (median: 12). Forty six percent of respondents had completed formal study to a year 12 (or equivalent) level while eighteen percent had not even completed formal study to a year 10 level.

Methods

To conduct this research the systematic methods used  to conduct a review of research on

Strategies were designed to assist low-achieving students. Where the research supported it, we used meta - analytic techniques. Otherwise, the reviews are narrative in form. In conducting the synthesis, they drew on guidance from previous researchers who have published on synthesis methodology and particularly from the work of Harris Cooper (1998).

Findings

Of the 125 respondents, 71 completed the course, 32 completed some components of the course and 22 withdrew without submitting any assessment tasks. The latter group were awarded a total course score of zero. This distribution is obviously not Gaussian and therefore it was inappropriate to apply standard linear regression models to the entire total course score distribution. The distribution of scores for students who completed all assessment items in the course, however, was close to symmetric, and we were able to apply a linear regression model to this subset of the data. The M-test results for all respondents ranged from 15 to 38 (out of a total score of 38) with a mean of 30. Only three students scored less than 19 in the M-test and all three subsequently withdrew from the course. Nineteen students scored less than 28 in the Mtest (approximately 75% of the total score) and of these, 14 subsequently withdrew from the course. These results indicate that students who achieved a poor score in the M-test were more likely to be at risk of withdrawing from the course; however many students who performed well in this test also withdrew or failed the course.

Implication

            In this article it was reported on an initial study that sought to assess the predictive validity of pre-course measures aimed to identify at-risk students in a distance education course. They used regression-type models for this purpose and were able to report on models of both student a chievement and completion. These models, however, were of limited use, as they both failed to explain very much of the variation in the final performance of students in our sample. It is possible that with the inclusion of more measures aimed to assess other areas in the affective and conative domains, we may be able 113 to develop a model that can explain more of this variation. Such a large number of items is counterproductive and in any case, with such a model we are only just over 50% certain that we can identify at-risk students.  The use of pre-course measures to identify at-risk students is akin to measuring height with a yardstick; they are of limited use. In this study we found some evidence to suggest that older students or those who perform poorly in a pre-course mathematics test are at risk of withdrawing from the course. Certainly many of the students who performed poorly in the M-test did not succeed in the course, however many students who did perform satisfactorily in the M-test also did not succeed. Despite the limitations of such tests, academics continue to use them, and direct students who perform poorly in these tests towards learning support programs. This is despite the measure not correlating with final performance when it is implemented early in the course and then moderately correlating with final performance when it is implemented mid-way through the course.