Samples and Variables
Our subjects to be measured are UBC undergraduate classes. There have been tens if not hundreds of thousands of classes offered at UBC since its establishment in 1915. Therefore it is wise to narrow down our population to Winter Session lectures from 2009 to 2013 because not only are older data not readily available, we also want to set a criteria that specifies certain characteristics of the classes we are interested in to ensure consistency amongst our sample in an effort to minimize the effect of confounding variables. First of all, we exclude lab courses and tutorials. These types of classes place an emphasis on individual tasks and are often run by teaching assistants. Secondly, we limit our sample classes to those offered only during the Winter Session since the summer semester spans a shorter period of time. Squeezing 3 months worth of work in under 6 weeks will have a profound impact on students’ performance. Lastly, we will only be surveying classes in the past five years, starting from 2009. This is because the structures of many UBC courses are being constantly modified. Classes are being added or dropped over the years, and many curriculums are still experimental. Five is simply an arbitrary number we have set in hope that the classes during this time period share more similarities than differences.
We will be looking at 2 numerical variables of our samples. The independent variable will be the initial class size, or the number of students who were registered in a particular section on day 1 of the semester. This figure will account for the number of students who withdrew before course completion. The dependent variable that could potentially be influenced by class size would be mean student score, which is simply the average score of those who completed the entire course.
Sampling Method
Because the UBC community consists of a diversity of academic programs, our study needs to address all the major disciplines that vary significantly amongst each other. Considering that faculties like Arts or Sciences offer a far wider selection of courses than Engineering or Music, a simple random sampling does not guarantee that our final sample includes courses that are representative of each subcategory. This problem can be easily resolved by employing stratified sampling. Moreover, this type of sampling process can help us observe any relationship that might exist amongst different academic disciplines. Therefore, we will classify all the undergraduate courses that UBC has offered during the past 5 years into five major disciplines: 1) Arts and Languages, 2) Science (including Faculty of Land and Food Systems, Forestry, and Human Kinetics), 3) Computer Science, Engineering and Mathematics, 4) Business and Management, and 5) Music and Visual Arts. The courses within these five strata should have considerable similarities.
The Office of Planning and Institutional Research (PAIR) releases numerous statistical reports online to the general public, like classroom spaces and degree profiles. We will be accessing “Grades Distribution” on its web server to obtain the data necessary for our study. By simply entering the year and session (eg. 2012 Winter) and subjects (eg. Microbiology), we will receive a complete spreadsheet of all the sections of all the courses with their respective data. We will then perform simple random sampling of the classes belonging to a specific discipline during the past five years to select 35 sections from all those that satisfy our criteria. After obtaining 175 samples from 5 strata, we will record the variables of our interest: initial class size and the average score.