diff --git a/ASEE-DELOS_Cooper.tex b/ASEE-DELOS_Cooper.tex index 9758522..6b679b6 100644 --- a/ASEE-DELOS_Cooper.tex +++ b/ASEE-DELOS_Cooper.tex @@ -181,7 +181,7 @@ with the same data set and submit reports graded with the rubric in Appendix A. Lab \#1 asks students to quantify differences in machining methods between band saw and computer numerical control (CNC) parts. Labs \#2-4 ask students to quantify differences between rational predictions using -analytical and numerical models and empirical measurements for static and +rational models and empirical measurements for static and dynamic cantilever beams. In the PjBL activity, the Lab \#5 competition, the students are given the task to create a design of experiments, create a predictive model, and use engineering judgment to measure the mass of an object @@ -229,7 +229,7 @@ experiments, take measurements, and create finite element analysis models. The competition does not have calibration weights, so the students have to rely on rational predictions and engineering judgments. The competition ends with the submission of their best estimate of object mass with -a propagation of error and the Methods section. The lab group with the most +a propagation of error and the lab report's Methods section. The lab group with the most accurate measurement is awarded a \$150-prize. After the prize is awarded, the actual object masses are announced. The lab groups use week 12 to revise their approach and submit the lab report. The goal is to encourage students to @@ -247,34 +247,33 @@ clear feedback on the final error in the predicted results. %-------------------------------------- %report 23.7442 4.0000 1776.0000 0.0000 %====================================== -The course focuses on problem-solving and technical writing. The F-value in a -one-way repeated Analysis of Variance, using the Python package -statsmodels\cite{seabold2010} was 23.74 between labs 0-4 with 445 -students indicating that there was a statistically significant affect on lab -report grades. In Fig.~\ref{quality}(a), the scores of each lab group is fit to -a linear model to determine the change in report grade per report between Labs -\#0-4. The goal was to have the entire class in the green ``continuous -improvement''-area. In Fall~2018, 56\% of the class continually improved and in -Fall~2019, 59\% of the class continually improved their scores. The ``maintain -quality'' area represents students that write reports of high quality initially, -but do not improve during the course of the class. In Fall~2018 and Fall 2019, -the students that maintained quality accounted for 43\% and 36\%, respectively. -The remaining 1\% and 4\% of the class did not improve or meet specifications -for lab reports, in Fall 2018 and 2019, respectively. The grades from Labs~\#5-6 -are shown in Fig.~\ref{quality}(b). Lab~\#5 was the PjBL contest and marked a -significant increase in expectations. The results of this study, suggest that -students are able to incorporate feedback from teaching assistants and myself -and show improvements in technical writing. The Labs increased in difficulty, so -even the groups of students that maintained their grade at the specified level -show marked improvement in communicating difficult concepts. - -Regarding the effectiveness of specifications grading in technical writing, -there is still a normal distribution of grades with the class mean between 80 -and 85~points and grades increased throughout the semester. One argument against + +The course focuses on problem-solving and technical writing. In Fig.~\ref{quality}(a), +the scores of each lab group is fit to a linear model to measure average increase in +grade per report between Labs \#0-4. The goal was to have the entire class in the green +``continuous improvement''-area. In Fall~2018, 56\% of the class continually improved and +in Fall~2019, 59\% of the class continually improved their scores. The ``maintain +quality'' area represents students that write reports of high quality initially, but do +not improve during the course of the class. In Fall~2018 and Fall 2019, the students that +maintained quality accounted for 43\% and 36\%, respectively. The remaining 1\% and 4\% +of the class did not improve or meet specifications for lab reports, in Fall 2018 and +2019, respectively. The F-value in a one-way repeated Analysis of Variance of lab report +scores, using the Python package statsmodels\cite{seabold2010}, was 23.74 between labs 0-4 +with 445 students indicating that there was a statistically significant affect on lab +report grades. The grades from Labs~\#5-6 are shown in Fig.~\ref{quality}(b). Lab~\#5 was +the PjBL contest and marked a significant increase in expectations. The results of this +study, suggest that students are able to incorporate feedback from teaching assistants and +myself and show improvements in technical writing. The Labs increased in difficulty, so +even the groups of students that maintained their grade at the specified level show marked +improvement in communicating difficult concepts. + +I found specifications grading in technical writing to be an effective method of +evaluation. The grades are normally distributed with the class mean increasing from 80 +to 85~points. One argument against specifications grading is that students may not be motivated to increase their grade because once the grade is above passing there is no incentive to improve. I find a clear increase in grades throughout the semester, and the students -that were in the ``maintaining poor quality'' regime did fail and redo lab reports. +that were in the ``maintain poor quality'' regime did fail and redo lab reports. The students that did not improve found great difficulty in Labs~\#5-6, most failing those assignments and revising their work. The specifications grading also has the most noticeable effect on under-performing students. The students @@ -321,7 +320,7 @@ progress was sustained and labs did not become more demanding. semester, Green indicates students that passed Report~\#0 whose scores continued to increase throughout the semester, and orange are students that passed Report~\#0, but their scores decreased throughout the semester. The orange marks - in the red sections, "maintain poor quality" were at risk of failing other lab + in the red sections, ``maintain poor quality'' were at risk of failing other lab reports. In (b), box plots of the scores from 2018 and 2019 on reports 0-6 are plotted. The median is shown by a horizontal line, the notches indicate the confidence interval, the whiskers denote the range of scores, with outliers @@ -335,31 +334,30 @@ histogram of errors based upon reported results demonstrate the range of effectiveness of each lab group's experimental work. In Fall~2018 and Fall~2019, the average and standard deviation in error to measure a 32-g object was 18.3$\pm$32.8~g and 11.4$\pm$26.7~g, respectively. While top three most accurate -reports had errors less than 4\%. The competition provides very specific -feedback to lab groups, and provides a non-grade-based metric to evaluate +reports had errors less than 4\%. The competition provides specific +feedback to lab groups, and a non-grade-based metric to evaluate student effort and learning. -This PjBL Lab qualitatively had the highest enthusiasm and participation from -the students. Student SET responses included, ``I liked the mass measuring -contest!'', ``I liked using ANSYS and the competition.'', ``I liked the -competition where the answer was unknown. I think that was the most beneficial -thing we did and I think more of those labs would be helpful.'' Attendance to -announce winners of the contest was not mandatory, but over 90\% of the class was -present. Students compared answers, studied methods, and results. After the -object masses were given to the class, they revised their methods one more time -to reduce errors in their data collection and processing. The benefit of the -contest was the increased enthusiasm in studying beam dynamics and finite -element methods. Even students that had very high errors, had finite element -models with demonstrated convergence, fast fourier transform analysis of natural -frequencies of cantilever beams. These competitions work best when the learning -happens whether or not the group wins\cite{burguillo2010}. +This PjBL Lab qualitatively had the highest enthusiasm and participation from the +students. Student SET responses included, ``I liked the mass measuring contest!'', ``I +liked using ANSYS and the competition.'', ``I liked the competition where the answer was +unknown. I think that was the most beneficial thing we did and I think more of those labs +would be helpful.'' Attendance to announce winners of the contest was not mandatory, but +over 90\% of the class was present. Students compared answers, studied methods, and +results. After the object masses were given to the class, they revised their methods one +more time to reduce errors in their data collection and processing. These competitions +work best when the learning happens whether or not the group wins\cite{burguillo2010}. The +benefit of the contest was the increased enthusiasm in studying beam dynamics and finite +element methods. Even students that had very high errors demonstrated finite element +models convergence and fast fourier transform analysis of natural frequencies of +cantilever beams. \begin{figure}[ht!] \centering \includegraphics[width=5in]{./track_progress/mass_measure.png} \caption{Plotted above is a histogram of the reported errors from Fall~2018 - and Fall~2019 for the mass measurement contest. The average mass reported in - Fall~2018 and Fall~2019 was 18~$\pm$~33~g and 41~$\pm$~27~g, respectively with + and Fall~2019 for the mass measurement contest. The average error in mass reported in + Fall~2018 and Fall~2019 was 18~$\pm$~33~g and 11.4~$\pm$~27~g, respectively with error reported as standard deviation. The actual mass measurements were 32~$\pm$~2~g. The histogram is the error=(reported value - the actual value). \label{contest}} \end{figure} @@ -409,18 +407,18 @@ over 50\% of students increased technical writing quality. Access to interactive notebooks increased the variety and use of the lab handouts. Using Jupyter notebook handouts created a medium that mixed background information, data processing, and simple engineering models. The Jupyter notebooks helped to close -the gap between rational, thinking design and empirical, hands-on design. The -project-based upper engineering lab course redesign has been a success. Using +the gap between rational and empirical, hands-on design. The +project-based upper-level engineering lab course redesign has been a success. Using the 2019-2020 senior capstone students, I found a statistically significant increase in preparation for engineering design from taking the lab course with PjBL. Some ongoing work will be to evaluate the effectiveness of individual changes in the course. Specifications grading is a novel way to asses engineering students' -technical writing skills. I believe the process of revising reports provided +technical writing skills. I believe the process of revising reports provides much-needed practice for students, but it would be interesting to see what fraction of the class has measurable increase in writing quality without this -process. I assume the PjBL competition was a big motivational and preparational +process. I assume the PjBL competition is a big motivational and preparational tool, but there may be other sources of motivation and preparation. Some future work is to compare results between a competition-based PjBL and PjBL component with no competition and to incorporate senior design grades into