Sample Size and Representation During Assessment

This article (http://pareonline.net/getvn.asp?v=9&n=11) published by Jason W. Osborne and Anna B. Costello from North Carolina State University I found interesting because the topic is something that I struggle with continually as I am sure most institutional researchers do.

Working in a smaller to medium size university gathering large sample sizes during assessments can be challenging. My previous postings discussed the CLA, they require a minimum sample size of 100 students per administration. In my perspective, finding 100 students willing to take a 90 minute writing assessment during regular class time has been somewhat of an issue. I have met the minimum requirement of 100 students, but narrowly.

The reason for a large sample size is to minimize the probability of errors and create the ability to generalize to the population at large. The assessments administered on the MSU campus are typically reported to the North Dakota University System for accountability reasons, these results are viewed by legislators and the State Board of Higher Education. Often times the results can lead to policy change and new criteria on campus....this is why the need for a solid study or assessment with a representative sample is necessary.

The author of the article notes that no sample is perfectly reflective of the population, but the higher the N the better the chances of representation. The author explains this notion by utilizing a number of variables to identify what form of mass would calculate the most return. The variables calculated were: subject-to-item ratios, variable-to-component ratios, extra matrices, and correct factor structure (conducted by factor analysis).

The major findings was that as mentioned previously size does indeed matter. The larger the N no matter what type of variables or analysis used.