Abstract: Context: Online video lectures have been available to students through platforms like YouTube for approximately 10 years. Content varies from full-length lectures that have been recorded during live lectures, to PowerPoint presentations with voiceover, to shorter segments recorded using a tablet and screen recording software. The statistical analysis of global online watch data of publicly-available engineering undergraduate content placed on YouTube offers an ability to examine how students from around the world use this resource. The variability in viewing behaviour provides information about both the students and the country from which the content is being viewed. This offers the opportunity to estimate and consider how this type of learning will grow with time in numerous countries around the world.
Purpose: To develop an understanding of how engineering students worldwide use publicly-available online video lectures and to draw inferences about how and why content viewing varies by both country and time.
Approach: From September 2013 to December 2015, three Mechanical Engineering undergraduate courses were developed for online delivery and uploaded to YouTube. The content included segments that vary in duration from 5 to 15 minutes, and were recorded using a tablet computer with screen recording software. This content has been used for blended / flipped delivery by the first author in his assigned teaching, but it has also been made available for public viewing at no cost. YouTube Analytics has been used to extract global online watch data including the number of views and number of watch minutes. Statistical analysis has been performed on this data in order to provide insight into how online content is used by students.
Results: Data analysis reveals annual cycles in watch data that correlate to both academic schedules and weekly cycles. Rank-ordered plots of watch data reveal a decay rate of -1.32 by country and -0.60 by video. A change in decay rate slope may be attributed to the coherence or completeness of the total data set being examined. In the case of country data, this is most likely due to English proficiency and the percentage of internet adoption. In the case of video data, this is most likely due to the multi-year staggered nature by which courses have been posted. A scaling law is proposed that relates annual watch minutes per video segment per country to inverse engineering enrolment raised to the power minus 4/3. The magnitude of this relationship is approximately dependent on English proficiency index and the percentage of internet adoption, although exceptions to this trend are noted for certain countries.
Conclusions: Global online watch data reveals both annual and weekly study habits of students. Rank-ordered power-law plots reveal a change in slope that corresponds to the coherence or completeness of the total data set. Average internet bandwidth, provided that it is above 500 kbps, was found to have no noticeable influence on viewing habits. A scaling relationship is proposed that is able to collapse watch minute data in relation to engineering undergraduate enrollment, English proficiency, and household internet adoption. It is anticipated that watch data in certain countries will increase as the percentage of households with internet increases or as restrictions on YouTube are relaxed.
To cite this article: Hugo, Ronald J and Meikleham, Alexandra. Statistical analysis of global online watch data [online]. In: 27th Annual Conference of the Australasian Association for Engineering Education : AAEE 2016. Lismore, NSW: Southern Cross University, 2016: 357-366.
[cited 28 May 17].