A freely available educational application (a mobile website) is presented. This provides access to educational material and drilling on selected topics within mathematics and statistics with an emphasis on tablets and mobile phones. The application
adapts to the students performance, selecting from easy to difficult questions, or older material etc. These adaptations are based on statistical models and analyses of data from testing precursors of the system within several courses, from calculus and introductory statistics through multiple linear regression. The application can be used in both on-line and off-line modes. The behavior of the application is determined by parameters, the effects of which can be estimated statistically. Results presented include analyses of how the internal algorithms relate to passing a course and general incremental improvement in knowledge during a semester.
In this paper the TileShuffle method is evaluated as a search method for candidate lncRNAs at 8q24.2. The method is run on three microarrays. Microarrays which all contained the same sample and repeated copies of tiled probes. This allows the coheren
ce of the selection method within and between microarrays to be estimated by Monte Carlo simulations on the repeated probes.
An on-line drilling system, the tutor-web, has been developed and used for teaching mathematics and statistics. The system was used in a basic course in calculus including 182 students. The students were requested to answer quiz questions in the tuto
r-web and therefore monitored continuously during the semester. Data available are grades on a status exam conducted in the beginning of the course, a final grade and data gathered in the tutor-web system. A classification of the students is proposed using the data gathered in the system; a Good student should be able to solve a problem quickly and get it right, the diligent hard-working Learner may take longer to get the right answer, a guessing (Poor) student will not take long to get the wrong answer and the remaining (Unclassified) apparent non-learning students take long to get the wrong answer, resulting in a simple classification GLUP. The (Poor) students were found to show the least improvement, defined as the change in grade from the status to the final exams, while the Learners were found to improve the most. The results are used to demonstrate how further experiments are needed and can be designed as well as to indicate how a system needs to be further developed to accommodate such experiments.
Research is described on a system for web-assisted education and how it is used to deliver on-line drill questions, automatically suited to individual students. The system can store and display all of the various pieces of information used in a class
-room (slides, examples, handouts, drill items) and give individualized drills to participating students. The system is built on the basic theme that it is for learning rather than evaluation. Experimental results shown here imply that both the item database and the item allocation methods are important and examples are given on how these need to be tuned for each course. Different item allocation methods are discussed and a method is proposed for comparing several such schemes. It is shown that students improve their knowledge while using the system. Classical statistical models which do not include learning, but are designed for mere evaluation, are therefore not applicable. A corollary of the openness and emphasis on learning is that the student is permitted to continue requesting drill items until the system reports a grade which is satisfactory to the student. An obvious resulting challenge is how such a grade should be computed so as to reflect actual knowledge at the time of computation, entice the student to continue and simultaneously be a clear indication for the student. To name a few methods, a grade can in principle be computed based on all available answers on a topic, on the last few answers or on answers up to a given number of attempts, but all of these have obvious problems.
Statistical multispecies models of multiarea marine ecosystems use a variety of data sources to estimate parameters using composite or weighted likelihood functions with associated weighting issues and questions on how to obtain variance estimates. R
egardless of the method used to obtain point estimates, a method is needed for variance estimation. A bootstrap technique is introduced for the evaluation of uncertainty in such models, taking into account inherent spatial and temporal correlations in the data sets thus avoiding many model--specification issues, which are commonly transferred as assumptions from a likelihood estimation procedure into Hessian--based variance estimation procedures. The technique is demonstrated on a real data set and used to look for estimation bias and the effects of different aggregation levels in population dynamics models.
