Presentation on theme: "Edmund Cannon Electronic and traditional learning on an econometrics unit September 2009."— Presentation transcript:
Edmund Cannon Electronic and traditional learning on an econometrics unit September 2009
2 Introduction Which students use Blackboard (BB) and how do they use it? Handouts (Word/pdf documents) versus Power-point slides. Possible consequences.
3 Estimating treatment effects Even with a perfectly-designed controlled experiment, effects would need to be large for them to appear statistically significant (sample size 145 or fewer). I have not performed a controlled experiment so any correlations will not identify treatment effects. But we can still learn something... Nb in idiot regressions, the best explanators for exam performance are lecture attendance and 1 st -year exam mark.
4 Powerpoint etc. Chen and Lin (2008, IREE) argue PP slides improves student performance: other papers more cautious. Students like PP. Need to think about how they use it.
5 What are print-outs of PP slides for? Probably helps dyslexic students. May reduce the number of mistakes on some of the pieces of paper in a students file. May make students more passive (substitutes visual for kinesthetic learning). May change a students lecture attendance. May substitute/complement other paper resources.
6 What are (Word doc) handouts for? Nb assuming a technical unit like econometrics Supplementary information to the text book. Provides material to discuss in the lecture/class. Lecturer too lazy / arrogant to use text book properly.
7 QM3 Econometrics Compulsory second-year unit for most Economics students Excluding Erasmus, etc, I analyse 145 students (whole sample) or 115 students (omit EconMath). All students have AAB+ at A-level, Maths A-level grade A/B and have passed QM1 (Maths) and QM2 (Statistics) in the first year. Most students have a further compulsory econometrics unit in the third year (not EconMath).
8 QM3 Econometrics, cont'd Textbook-based unit testing technical skills (Murray: Econometrics: A Critical Introduction). First term overlaps QM2 material - OLS (chapters 1-10). Second term all new – heteroskedasticity, auto-correlation, consistency, IV/2SLS, Demand and Supply (chapters 11-14). Exam: answer three questions on algebra, stata, economic interpretation.
Statistics for total BB useage 9 QM3QM2 / Stat EverTerm 2VacTerm 3EverTerm 2VacTerm 3 I II.i II.ii III F Ordered probit relating class of mark to BB use is significant [p = 0.018].
10 What did I put on Blackboard? Exercises (one electronic, the rest paper). Handouts (Word/pdf documents). Power-point files for lectures. Lecture given PP slides put on BB Word doc put on BB Lecture 1530 Jan7 Feb Lecture 166 Feb22 Feb Lecture 1713 Feb22 Feb Intro to IV16 Feb Sargan16 Feb
Number of times students accessed teaching material 11 Tentative conclusion: students use BB as a mobile reference
Number of times accessed this doc on BB 12 QM3Power-point slides (L15)Handout (Intro IV) EverTerm 2VacTerm 3EverTerm 2VacTerm 3 I II.i II.ii III F Nb strong correlation between downloading handout and downloading PP
Number of times accessed 13 QM3Power-point slides (L17)Handout (Sargan) EverTerm 2VacTerm 3EverTerm 2VacTerm 3 I II.i II.ii III F
14 First time students accessed PP slides
First time students accessed handout 15
Proportion of students accessed on BB for the first time 16 QM3Power-point slides (L17)Handout (Sargan) EverTerm 2VacTerm 3EverTerm 2VacTerm 3 I II.i II.ii III F Conclusion: many students did not access material that was needed to supplement the textbook.
17 Correlation with exam performance QM3 mark Chen-Lin Attend L (6.5) 18.6 (9.6) [0.029] PP L (8.5) 17.4 (10.8) [0.0149] Attend * PP-19.8 (8.8) (10.7) [0.0316] Lecture att2.3 (1.2) 1 st year mark4.9 (1.3) N R-squared Panel
18 Conclusions Positive correlation between using BB and exam mark – tells us nothing new; Students access material on BB repeatedly (reference); Both good and bad students access material after course is over (consistent with cramming); What are students using material for?
19 Estimating treatment effects Consider a perfectly-designed controlled experiment, where the treatment effect is a binary variable and the explained variable is exam mark. Suppose the cross-section regression has an R 2 of 0.15 (which would be considered reasonable), that the standard deviation of exam mark is 15% and the treatment effect raises performance by 5%. Degrees of freedom must be about 150 for statistical significance.