STUDENT EVALUATIONS OF MODSTAT

A paper by Joseph M. Nowakowski - - Muskingum College

Making Statistics More Effective in Schools of Business "Technology, Learning and Statistics" 12th Annual Conference - June 26-28, 1997 The Pennsylvania State University

STUDENT EVALUATIONS OF MODSTAT
Joseph M. Nowakowski, Muskingum College
132 Cambridge Hall, New Concord, OH 43762
INTRODUCTORY STATISTICS, SOFTWARE

Recognizing the necessity of statistical analysis as one of the tools successful job seekers bring to today's workplace, most business departments offer one or more courses in statistics as either required or highly recommended elective courses. Given the enhancement of computing power, many instructors choose to incorporate some computer--based assignments in their classes. In many cases, especially at larger colleges and universities with higher levels of computer accessibility, the introductory statistics course has evolved into a lecture/lab course, where a substantial amount of course work is devoted directly to computer instruction and hands-on operations. In smaller schools, lack of adequate computer facilities or software has prevented, in some cases, the evolution toward computer-based classes, even when desired by both faculty and administration. Even at larger schools, a problem exists in the undergraduate statistics course concerning the choice of software. In a semester long course, it would obviously be impractical to try to introduce students to a wide variety of statistical software programs. Selection of one program invariably entails an opportunity cost as certain features of other programs are sacrificed.

In addition to the possibility of inadequate resources and the problem of software selection, a further concern is that of how well courses prepare students for future class- or work-related use of statistics. It may well be that, after spending hours in the computer lab gaining some familiarity with a particular statistical package, the individual enters another course or a job where statistical analysis is done on another, unfamiliar, software program. Although an argument can be made that gaining experience with the first program places one higher up the learning curve, the frustrated student or employee may fail to see the benefits of having learned Package A when s/he must learn to use Package B, with the prospect looming of learning Package C in the next course or position.

Given these considerations, some instructors may find a non-networked, self-contained, statistical package an attractive option. A 'walk-away' package has the advantage of enabling the individual to use it upon entering the next class or job. To be an attractive tool for the student, however, the package must be fairly easy to use; powerful enough to perform a variety of analyses, many beyond that of the typical introductory course; and priced affordably.

MODSTAT is a self-contained statistical software package designed for introductory and advanced statistics students. The software was developed by Dr. Robert C. Knodt (University of Richmond, retired) for Modern Microcomputers. The package contains a large number of statistical routines in a menu-driven (non-Windows) format. One important feature of the software is its inclusion of suggestions for the proper statistical routine to perform in different situations. An added attraction of this feature is the concurrent introduction to previously unknown techniques which may foster greater interest in the student.

Three obviously attractive features of MODSTAT are portability, power, and ease of use. The individual version of the software comes on a single disk. The program is powerful, with over 180 routines, but inexpensively priced at $22.00 for personal copies. There is also a network option available. The available routines include applications that mimic what one would expect to find in any introductory statistics course. An examination of available routines, part of a determination whether the software would be useful with the business statistics course as currently taught at Muskingum College, yielded possibilities for at least weekly assignments with no changes in the syllabus other than introducing the software to the students.

In a statistics course without a lab, many professors desiring to increase the use of computers may find themselves forced to reduce lecture time more than is wished. The more demonstration and hand-holding the students requires, the less time is available for other classroom needs. Additionally, the less comfortable students feel with a program, the more resistance is likely to he encountered. Therefore, an important element in a software program is ease of use.

Ease of use was evaluated by asking introductory statistics students, without prior exposure to the software, to perform a series of elementary tasks. These tasks would be included in most introductory business statistics classes: finding descriptive statistics for a small data set, constructing a confidence interval, identifying outliers, calculating various probabilities, translating z-scores into raw scores and vice versa.

The students kept track of the time needed to complete each task, and then completed a survey concerning their use of the software. The results of the students' evaluations arE summarized to give potential users an idea of how accessible this software would be to students. Table 1 below summarizes the students' performance on the assignments and their reactions to the software.

The students were given two weeks during which they could sign up to use the software and earn the extra credit. Early users did not seem to coach later users: there is no trend in the times of completion for any of the tasks over the two week period.

Students were told that the results would be used in a summary to be distributed to other interested people. They were also given extra credit for participation, and were told that they would be evaluated based on completion of the survey, not on perfect completion of the assignments. Working individually, they carried the instruction sheet, including the survey, to an empty office with a single computer. The only general instructions were:

Turn on the computer. At the C:\ prompt, type cd c:\modstat and hit return, then type menu and hit return. Now you're on your own.

The survey asked them to complete four general types of assignments. The first involved calculating descriptive statistics (mean, standard deviation, median, interquartile range) on a small set of stock returns, estimating a 95% confidence interval on the stock values, and identifying any outliers. Each student examined a different set of ten stocks. The stock values had to he entered into MODSTAT's data base and then analyzed. The mean time of completion was under twelve minutes. A high outlier (>2.5 standard deviations from the mean) raising the mean was detected by MODSTAT. The median was ten minutes.

Next the students were asked to transform the stock values into standard normal variables. This procedure was not part of the main menu in MODSTAT, but rather a sub-routine offered after the descriptive statistics had been calculated. Several students failed to complete this task; those who did averaged slightly under twelve minutes. Again, a high outlier was detected. The median time was ten minutes.

The third assignment directed the students to complete a probability table for a Poisson distribution, with which the students were familiar. Lambda was based on the number of letters in the students' last names, so very few students had the same value for lambda. Mean time of completion was slightly over seven and one-half minutes. The median was five minutes. Lilliefor's test for normality suggested a skewed distribution (a=0.01).

In the final software assignment students had to find various areas under a normal curve. The mean time of completion was just over nine minutes, with a median time of seven minutes. There was one high outlier detected. This was, conceptually, probably the most difficult task, but it was the second most quickly completed task on average.

The total time students took to complete all four assignments had a mean of slightly under forty-two minutes. The distribution of times was non-normal at the 5% level, but not at the 1% level. The median time of completion was forty minutes.

Given the fact that there is no computer lab work as part of the statistics class in which these students were enrolled, completing these assignments in under an hour (in most cases) with no real instructions or help beyond that in the program suggests that the software is fairly easy to use. Obviously, some introduction and guidance from the instructor would have reduced the time needed by the students.

Somewhat surprising was the generally negative reaction to the software by the students. The survey questions are reproduced below. See also Table 2 for a summary. The relatively low scores from the survey reflect in part the students' low level of computer experience. Several students reported frustration using the software, but that would probably be reduced considerably were the students given some introduction to the program before being thrown into the fire. A casual observation is that students with more computer experience found the software quite easy to use.

Next year, MODSTAT will be introduced into the business and economic statistics class at Muskingum in a more involved trial. Students will be required to purchase individual copies and use the software throughout the semester. The hope is that they will find the software flexible and easy to use, making them more receptive to the use of statistical analysis in other courses and in the future.