Master's thesis: "An investigation in introducing Matlab and data analysis in introductory physics"
In my thesis I investigated if the use a certain media had an impact on first year physics students' learning outcomes. Statistical inference was applied to investigate this. The tests performed on the aquired data sets were: One-way ANOVA, G-test, chi square and McNemar's independent test, and lastly Student's paired t-test. Also some basic power calculations were performed in regards to the validity of the results of the statistical tests.
In 2009 I began my physics education at the University of Copenhagen. In connection to the laboratory
part, of the very first attended course Mek1 on introductory physics, we were to use the CAS-software Matlab
and apply error analysis with little instruction herein. Personally, this gave rise to much frustration.
This experience inspired the idea to author teachings materials on the basic use of Matlab and data analysis
to help future students avoid the frustration I felt. The teaching materials authored were a book titled
“Grundlæggende Matlab og dataanalyse” (the book) and in total 11 screencasts: Both conveyed the same in
terms of basic Matlab but the screencasts did not present any statistical background.
The main question we investigated in this thesis was whether students learned more using either the
book or the screencasts; but also if preference of a certain media in general was more efficient at schooling
purposes relative to others; here we looked at books, screencasts/videos, and notes on the internet in general.
In addition we also asked them whether they found either the book, screencasts, or the internet most
To measure learning outcomes three tests were given: A self authored test on basic Matlab and data
analysis (Lab Test), the Data Handling Diagnostics test (DHD), and lastly the Force Concept Inventory test
(FCI). The DHD was solely on both basic and advanced data analysis and the FCI tested for skills in Newtonian
mechanics; mostly misconceptions. Overall we found that the students did quite well on the three
tests as µLab Test = 14.92 ± 0.17, µDHD = 10.10 ± 0.23, and a gain in the FCI of 3.82 ± 1.01. The gain of the FCI
was found to be significantly greater than zero; as such we found statistical proof that the students did learn
something about Newtonian mechanics by attending the course Mek1. In terms of statistical concepts we
found that future courses of Mek1 should give more attention to SDOM whereas the spread and arithmetic
mean warrants no further attention. In addition we found that learning two to different CAS-softwares, here
Matlab and Maple, have given rise to confusion for the students.
The first four questions of Lab Test asked the students about their media preference. From these we
found that the book was best received as 48.89% of 136 students used it, and of the those 65.15% found it
the most educational. The screencasts where not very popular: Here only 28.68% used them and of those
25.64% also found them most educational. We argued that the durations of the screencasts were too long
and need trimming and revision to become a viable future resource. Additionally we found evidence by
coupling the students’ media preference to the other tests that the book along the teachings of Mek1 significantly
improved the students’ ability to use the error propagation formula and increased their knowledge
of precision of a physical quantity.
We on the other hand did not find evidence that the preference of a general media significantly increased
the mean test scores of either tests; also that the number of total correct and wrong answers (NTC) of either
tests was independent of general media preference. The book and the screencasts was also not found to
improve the overall mean test score and they were also independent of the NTC.
Lastly, we investigated what the term “data analysis” entails in the Danish secondary: Using the written
exams problems from STX and HTX we found that transformation, plotting and fitting of data were common
in both educations. In addition we gave a rough sketch as a suggested solution to introduce error analysis in STX and HTX which we dubbed “Error analysis bootcamp”.