Guiding students onto a successful undergraduate path. Are we taking the route through the “mole”-hill or are we climbing the mountain? David Booth, Nicholas Brewer, Linda Morris and David Coates School of Life Sciences: Learning & Teaching, University of Dundee, Scotland Abstract Prior to the development of a new Life Sciences curriculum at the University of Dundee, which began in the academic year 2011/12. It was clear that some students were weak in the applied numerate aspects of biosciences, which presented serious difficulties at the later stages of their degree. In order to mitigate this, going forward, in addition to raising entry requirements a system of bootstrapping was introduced. All students entering the undergraduate degree programmes in Biological/Biomedical Sciences undertake a ‘skills’ audit to assess their proficiency in maths, physics and chemistry. Those not achieving a satisfactory passing grade of 70% are automatically enrolled on a 20 credit intensive foundational module that supports their transition to independent learning and enhances their integration into the theory and practical modules that run parallel. This module takes place over 10 weeks and contextually covers topics such as diverse as moles/molarity, use of logs and the physics of fluids. As the University of Dundee values informed teaching through an empirical approach where possible, we explore the effectiveness of this module. In a post-hoc, non-invasive and multidimensional approach, we analyse the strengths and weaknesses of these students relative to their peers. Data is derived from the pre-entry skills audit, in-course assessments across shared modules and running average grades from four cohorts of students entering first year between 2011/12 and 2014/15. We have evidence that students are embracing the module and are integrating into promising scientists. At the half-way point of their academic career at the University of Dundee their performance is indistinguishable from the rest of the cohort. Author keywords: PCA, Numeracy, Skills Audit Introduction Key STEM skills and knowledge can be often overlooked and under-resourced in the morass of subject specific material. As an example though it is an intrinsic component of the scientific method, statistical analysis is widely recognised as both difficult to teach and the boon/bane of the undergraduate student scientist (Reid and Petocz 2002). Such skills are in high demand and it is universally recognised in the literature that ability to conduct analysis improves the depth of a science students understanding (Colon-Berlingeri & Burrowes, 2011); permits the development of models (Elser and Hamilton, 2007); provides a strong foundation for quantitative approaches to laboratory work (Metz, 2008); organize data (Rein et al. 2007); and rationally evaluate the strength of arguments (Schield, 2004). Given the highly competitive nature of employment in the sciences, both academia and industry, it is vital to equip young scientists with the necessary prerequisites skills and knowledge for work. It is a matter of easily ascertainable fact that the generic STEM skills and knowledge are of most value to graduates in the life sciences at present, with many postgraduate and advanced courses pay particular attention to fostering these skills precisely because of the positive effect it has on their graduates in terms of employability or recruitment into doctoral training programmes. This position is further compounded by the substantive differences of scale and complexity between 20th and 21st century science. Those that possess the skillset talk the unbeaten path, and wrestle with big data can command the lion’s share of insight and prestige, a notion reinforced by the results of a recent survey conducted by Dice (2014). Moore (1997) argues that sciences with a strong mathematics component need no longer be deliberately inaccessible and that it should be a common aim to have basic quantitative literacy for all students in the 21st century. Though crucially and directly related to this work students of the biosciences are moving into a field of study that is increasingly driven by technological innovation and rapid refinement/evolution of technique and theory. As such the prerequisites of student skills and knowledge are dramatically different from those of previous decades. Such a sea-change in the style of research that is now both high-throughput and deeply computational (research areas that are often suffixed with -omics) is evidences in the multidisciplinary teams focused on the field of synthetic biology (EASAC policy report 13, 2010), and established fields of drug discovery (Sun et al. 2013), personalised medicine (Everett et al. In Press) and bioinformatics (Ouzounis 2012). Beginning in the academic year of 2011/12, a new curriculum in the Life Sciences was implemented at the University of Dundee. The recent extensive realignment and redesign of the University of Dundee life sciences undergraduate curriculum presented a suitable blank canvas with regards to the teaching and assessment of mathematics, chemistry, physics and statistics. The new ethos of the pedagogical approach is one engrained in the teaching of key skills; critical thinking; interactive workshops; and training in “state of the art” techniques. Prior to this, an audit and subsequent checklist of the skills and knowledge believed to be core to all Life Science students prior to entry into research led teaching was mapped out. One particular area of noted from previous cohorts was a weakness in not just basic scientific numeracy but the ability to apply factors, convert units, calculate molarity and perform basically logarithmic conversions and interpretations. For some even the basic organic chemistry was identified as an issue. Though there are a penumbra of module or instructor specific elements of much less formal non-biological STEM material delivered, according to the perceived needs of the particular curriculum element in question, the student experience and knowledge delivered would vary wildly with the associated instructor. Feedback and consultation with teaching staff who directly supervise honours students indicated that the “better” students at level 3 and beyond retain a broad awareness of the the different basic STEM knowledge and techniques, but that significant misconceptions still linger, and that the “weaker” students absorb almost nothing of value. To address this a module entitled BS11005 - An introduction to Maths, Physics and Chemistry was developed for those that may have identifiable weaknesses in areas described above. Prospective students obtaining the entry requirements for a degree would be offered a place on the programme, and subsequently be identified as individuals that may benefit from enrolment to this inductor module during this key transitionary period. A skills audit comprised of questions selected at random from the following topics: Basic algebra, reading graphs, basic inorganic chemistry, log calculations, moles and molarity, basic organic chemistry, basic physical chemistry, electrical principles, semi-log graphs and statistics was developed. Students were presented with thirty questions drawn from this question bank. On submission of the test students were informed as to whether they had achieved a satisfactory score or if they should try again. The composition of the skills audit would variable slightly per test such that it would be approximately 11-13 Mathematics 2 questions, 3-4 statistics questions, 10-13 chemistry questions and 3 physics questions. The audit had a time limit of 45 minutes and automatically submitted the answers after that time. The pass mark was arbitrarily set at 70%. The skills audit opened one month prior to matriculation with the advisor of studies in early September. The module ran for 11 weeks and covered topics in numeracy, physics of fluids, electrical principles, and fundamental physical and organic covered in two or three week blocks. Composed of a single lecture (two hours in duration) per week with five associated practical sessions, and three workshops. The module was assessed by five computer assessments, each worth 8% of the overall module mark and at the end of the module a one hour class test that covers all the material and was worth 60% of the overall module mark. Key questions that arise related to this approach are: 1. Can a skills audit prior to matriculation suitably discriminate those needing additional credits associated with mathematics, chemistry and physics? 2. Does this approach have a meaningful impact on the students exposed to the induction module? Are students completing the module academically competitive relative to their peers? Methodology To explore the discriminatory power of the skills audit 1102 responses from cohorts of students entering between the 2011/12 and 2014/15 academic years were collated. Individual questions within the audit were clustered according to subject marks received linearly transformed such that each correct response was worth a single mark. As students would answer a random subset of questions, a count of questions attempted per subject was made and the student performance in that area recorded as the proportion of correct responses. This developed a score for each of the four areas covered, namely Chemistry, Physics, Mathematics and Statistics. Those scoring less than 70% were classified as a “Fail”, and those scoring more classified as a “Pass”. To determine whether the composition of the audit in terms of question number per subject affected the student classification, a generalised linear model (GLM) with Poisson distributed errors was performed against question count for each subject set. To explore the relative contributions of the subjects towards producing a final classification, principal component analysis was performed with the first and second principal components correlated against the proportion of correct answers per subject. To explore the impact of BS11005 on subsequent performance, data for was collected in the form of a set of summative running average grades per student, across the core curriculum modules from those entering between the 2011/2012 academic year and 2013/2014 academic year. Students were anonymised and selected randomly from pools of those classified as failing and passing the skills audit (final n=100 for each group). Data was explored for tendencies and correlations; a one-way ANCOVA was conducted to determine a statistically significant difference between students attending BS11005 due to gaining a “Fail” vs “Pass” at skills audit on matriculation on the second year running average grades, controlling for first year running average grades; overall student performance was explored using one-way ANOVA and principal component analysis (Gorsuch 1983). Plotting, summary statistics, univariate and multivariate analysis were conducted using R (R Core Team, 2014). Analysis and discussion of skills audit Number of questions attempted in each component of the skills audit was found to not have a significant effect on the outcome of the test (GLM for each of the four subject sets p>0.05). However it was clear from principal component analysis that taken in an omnibus fashion the subjects contributed differentially towards classifying students as either a pass or fail (table 1). 3 PC1 0.285 Stat.score 0.447 Chem.score 0.327 Physics.score 0.782 Standard deviation 0.328 Proportion of Variance 0.482 Cumulative Proportion 0.482 PC2 -0.180 -0.772 -0.181 0.582 0.254 0.289 0.771 Math.score PC3 0.551 -0.452 0.666 -0.221 0.184 0.152 0.923 PC4 0.763 -0.023 -0.646 0.005 0.131 0.077 1.000 Table 1. Varimax rotated component loadings for four variables derived from the skills audit. An examination of the Kaiser-Meyer Olkin measure of sampling adequacy suggested a factorable sample (KMO = 0.68). All four subjects load positively onto the first principal component, with only performance in physics loading onto the second component (figure 1). This produces a characteristic striated effect in the analysis, with linear clusters (running bottom left to top right) based on physics performance, and those performing well in the other subjects tending towards the bottom right within these clusters (figure 2). 4 Figure 1. Explanatory variables plotted against first and second principal components. Indicating Students scoring highly chemistry, statistics and physics being loaded positively on the first principal component. 5 Figure 2. First and second principal components of 1102 skills audit attempts between 2011 and 2014. Explained variation of each component labelled on axis; attempts classified as fail coloured red those classified as a pass coloured blue, with confidence ellipses; arrow vectors indicating the loading of the four variables. From these data it would appear that whilst the skills audit is discriminating students based on ability, the quality of the assessment across the four key components is heterogenous. Assessment of chemistry, mathematics and statistics is functioning well, however physics delivering the least number of questions is effectively remaining untested. As such a more evenly constructed and granular form of assessment of student ability may be necessary to determine if a student requires remediation via BS11005. Likewise it may even be appropriate to produce a four point score for each student for personal reflection and the focusing of module choice out with the College of Life Sciences. Analysis of student performance post attendance BS11005 Students classified as failing during skills audit were found to differ significantly from those passing, however this difference between the two groups was not significant by the end of the second year (see figure 3). In exploring the impact of BS11005 on the trajectory of students into the research led teaching component of their undergraduate education, the second year running average was found to be significantly affected after controlling for first year performance. 6 Figure 3. Running average grade (mean ± 95%ci) of students passing in first year (18.1 ± 0.227) and second year (17.5 ± 0.298); and students failing the skills audit in first year (17.7 ± 0.278) and second year (17.5 ± 0.302). Asterisk above first year barplot indicates significant difference in the running average grade, using one-way ANOVA (F(1,198) = 6.562, p = 0.0112) with an effect size of 0.033. Figure 4. Running average grade of students passing in first year versus and second year. Students failing skills audit and taking BS11005 coloured red, those passing coloured blue. Though gradients were not significantly different, there was a weak but highly significant effect on the intercept, F(1,197) = 6.9493, p = 0.009) with an effect size of 0.03. Analysis of the student running average data set by component analysis separated individuals by performance across all modules with strongest students being negatively loaded on the first principal component (see figures 5 and 6). Modules have themes, focusing on different skills and knowledge and this is evidenced by the vectors of the factors loading in an almost perpendicular fashion between the first and second principal component. This is reinforced by strong correlation of first and second year running average. 7 Figure 5. First and second year running average grades plotted against first and second principal components. Indicating Students scoring highly in both years loading negatively on the first principal component. Figure 6. First and second principal components of 200 student running average grade sets between 2011 and 2013. Explained variation of each component labelled on axis; skills audit attempts classified as fail with BS11005 as an intervention coloured red those classified as a pass without BS11005 as an intervention coloured blue, with confidence ellipses; arrow vectors indicating the loading of the variables. Kaiser-Meyer Olkin measure of sampling adequacy suggested a suitably factorable sample (KMO = 0.84). Notably students classified as failing or passing the skills audit were not found to cluster together with both sets distributed almost equally distributed through the first and second principal components. Whilst the confidence ellipse of the failing students is larger than that of those passing, the two groups are indistinguishable from one another. 8 Discussion The use of a multivariate approach to exploring appears to be a valid tool for gleaning some insight into assessment validity and student performance; this approach has been well documented in the literature and applied to a variety of student cohorts and subjects (Gardner, 1972; Sullivan, 1996; Divjak and Oreški, 2009; Erimafa et al. 2009). Whilst these results are reporting observations, and not the outcome of a controlled trial, it seems plausible that intervening with an introduction to basic science would have a positive effect on students perceived as lacking those skills/knowledge. It is notable that both sets of students converge academically by second year. Whilst it was generally observed in CLS that high performing students can appear to excel across the board, there is a diverse range in student ability whereby those classified as weak or failing the skills audit at entry excel overall in the core curriculum modules. This lack of a relationship between student ability at matriculation might appear that to invalidate such a measurement as having poor validity in the long term, however it is clear that students can be effectively discriminated with careful selection of questions related to essential skills and knowledge required to fully embrace the core curriculum content. Student performance overall correlates strongly between level 1 and level 2, however It is interesting however to note that ability within one does not correspond with ability across the other, reflecting the multidimensional nature of student abilities and interests and the multidimensional nature of instructor assigned grades. Bowers (2011) noted with secondary school students that whilst instructor assigned grades can be subjective, that standardisation in core assessments can account for academic ability whereas instructor assigned grades act as a benchmark for motivation, attitudes and behaviours. Likewise Storkel (2012) reported diversity of ability and interests in pathology graduates enrolled to become clinicians. Given that students are receiving grades and assessment from a plethora of instructors, it seems unlikely that this would be a source of bias. References Anderson, L. W., Krathwohl, D. R., and Bloom, B. S. (2001) A Taxonomy for Learning, Teaching, and Assessing a Revision of Bloom’s Taxonomy of Educational Objectives, New York, NY:Longman. Brady, L. (1995) Curriculum development, 5th edn. Sydney, Prentice-Hall. Biggs, J. (1999) Teaching for Quality Learning at University. SRHE and Open University Press, Buckingham. Bowers, A.J. (2011) What's in a grade? The multidimensional nature of what teacherassigned grades assess in high school. Educational Research and Evaluation: An International Journal on Theory and Practice. Volume 17, Issue 3, 2011 Colon-Berlingeri, M. and Burrowes, P.A. (2011) Developing a Test of Scientific Literacy Skills (TOSLS): Measuring Undergraduates' Evaluation of Scientific Information and Arguments. CBE Life Sci Educ. 10(3): 259–267. 9 Crawford, C., Dearden, L. and Greaves, E. (2013) When you are born matters: evidence for England. IFS Report 2013.0080. London: Institute for Fiscal Studies. Crawley, M.J. (2007) The R Book. John Wiley & Sons, Ltd DelMas, R., Garfield, J., Chance, B., & Ooms, A. (2006) Assessing Students‘ Conceptual Understanding After a First Course in Statistics. Statistics Education Research Journal, 6(2), 28-58 Dice Salary Survey (2014) accessed at http://marketing.dice.com/ on the 1st Feb 2014. Divjak, B. and Oreški, D. (2009) Prediction of Academic Performance Using Discriminant Analysis. Proceedings of the ITI 2009 31st Int. Conf. on Information Technology Interfaces. Elser JJ, Hamilton A (2007) Stoichiometry and the New Biology: The Future Is Now. PLoS Biol 5(7): e181. doi: 10.1371/journal.pbio.0050181 Erimafa J.T., Iduseri A., and Edokpa I.W. (2009) Application of discriminant analysis to predict the class of degree for graduating students in a university system. International Journal of Physical Sciences Vol. 4 (1), pp. 016-021 Everett, Jeremy, Loo, Ruey Leng and Pullen, Francis S. (2013) Pharmacometabonomics and personalized medicine. Annals of Clinical Biochemistry. ISSN 0004-5632 (Print), 17581001 (Online) (In Press) (doi: 10.1177/0004563213497929 ) European Academies Science Advisory Council (2010) Realising European potential in synthetic biology: scientific opportunities and good governance. Policy report 13. ISBN: 9783-8047-2866-0 Fry, H., Ketteridge, S., and Marshall S. (2009) A Handbook for Teaching and Learning in Higher Education, Routledge, Abingdon Gardner, B. (1972) A multivariate computer analysis of students performances as a predictor of performance as a surgical intern. Journal of Surgical Research Volume 12, Issue 3, Pages 216–219 Garfield, J. (2006) Collaboration in Statistics Education Research: Stories, Reflections, and Lessons Learned, in International Statistical Institute Proceedings of the Seventh International Conference on Teaching Statistics [online]. Available at www.stat.auckland.ac.nz/~iase/publications/17/PL2_GARF.pdf. Garfield, J.B. and Ben-Zvi, D. (2007) Developing students' statistical reasoning: connecting research and teaching practice. Emeryville, CA: Key College Publishing. Gorsuch, R.L.(1983) Factor analysis (2nd ed.) Hillsdale, NJ: Erlbaum Hammersley, M. and Traianou, A. (2012) Ethics and Educational Research, British Educational Research Association on-line resource. Light, R. J., Singer, J. D., & Willett, J. B. (1990) By Design: Planning Research on Higher Education, Cambridge, MA: Harvard. Lock, R., Salt, D. & Soares, A. (2011) Subject Knowledge and pedagogy in science teacher training. Education research [online]. Available at http://www.wellcome.ac.uk/About-us/Publications/Reports/Education/ 1 0 Loughland, A., Reid, A., and Petocz, P. (2002) Young People’s Conceptions of Environment: A Phenomenographic Analysis. Environmental Education Research, 8, 187197. McGowan H. M. (2011) Planning a Comparative Experiment in Educational Settings. Journal of Statistics Education, Volume 19, Number 2 Metz, A.M. (2008) Teaching Statistics in Biology: Using Inquiry-based Learning to Strengthen Understanding of Statistical Analysis in Biology Laboratory Courses. CBE Life Sci Educ. 7(3): 317–326. doi: 10.1187/cbe.07-07-0046 Moore, D. S. (1997) New Pedagogy and New Content: The Case of Statistics. International Statistical Review, 65: 123–137. doi: 10.1111/j.1751-5823.1997.tb00390.x Morgan, B. (1999) "What are we Looking for in Theological Reflection?" Ministry Society Theology, 13(2), 6-21. Ouzounis, C.A. (2012) Rise and Demise of Bioinformatics? Promise and Progress. PLoS Comput Biol 8(4): e1002487. doi: 10.1371/journal.pcbi.1002487 Paranjape, M.D. (2010) Crisis of statistics pedagogy in India. International Association of Statistical Education (IASE) ICOTS8 Contributed Paper Refereed. Prideaux, D. (2003) British Medical Journal. 326:268-279. Print, M. (1993) Curriculum development and design, 2nd edn. Sydney, Allen & Unwin. R Core Team. (2014) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria. Http://www.R-project.org Ramsden, P. (1992) Learning to Teach in Higher Education, London: Routledge. Reid, A. and Petocz, P. (2002) Students’ Conceptions of Statistics: A Phenomenographic Study. Journal of Statistics Education Volume 10, Number 2 Reid (1997) "The Hierarchical Nature of Meaning in Music and the Understanding of Teaching andLearning," Advancing International Perspectives, 20, 626-631. Rein, D.C., Sharkey, J. and Kinkus, J. (2007) Integrating Bioinformatic Instruction Into Undergraduate Biology Laboratory Curriculum. Association for Biology Laboratory Education (ABLE) 2006 Proceedings, Vol. 28:183-216 Schield, M. (2004) Curricular Development in Statistics Education. 2004 IASE Roundtable, Lund Sweden Storkel, H. L. , Woodson, M. B. , Wegner, J. R. & Daniels, D. B. (2012) Multidimensional Student Assessment. The ASHA Leader. Sullivan, W.G. (1996) Multivariate analysis of student performance in large engineering economy classes. Frontiers in Education Conference. 26th Annual Conference., Proceedings of (Volume:1 ) 1 1 Taylor, C., Rees, G. and Davies, R. (2013) Devolution and geographies of education: the use of the Millennium Cohort Study for ‘home international’ comparisons across the UK. Comparative Education, 49(3), 290-316. Vere-Jones, D. (1995) The coming of age of statistical education. International Statistical Review, 63, 3-23. Xiaochen Sun, Santiago Vilar, and Nicholas P. Tatonetti (2013) High-Throughput Methods for Combinatorial Drug Discovery Sci Transl Med 2 [DOI:10.1126/scitranslmed.3006667] 1 2