The development of methods suitable to tackle the problem of inductive logic – how to justify arguments that generalise findings from data – has been signified by great controversies in the foundations and – later – also in statistics education. There have been several attempts to reconcile the various approaches or to simplify statistical inference: EDA, Non-parametric statistics, and the Bootstrap. EDA focuses on a strong connection between data and context, non parametrics reduces the complexity of the model, and Bootstrap rests solely on the data. Informal inference subsumes two different areas of didactic endeavour: teaching strategies to simplify the full complexity of inference by analogies, simulations, or visualisations on the one hand, and reduce the complexity of inference by a novel approach of Bootstrap and re-randomisation. The considerations about statistical inference will remain important in the era of Big Data. In this paper, the various approaches are compared for their merits and drawbacks.

Interactive dynamic technology can help students build concept images of core statistical ideas, in particular, ideas related to distributions and sampling, enabling them to make inferences from data and in the process help address common misconceptions about these ideas. Applet-like documents allow students to take meaningful statistical actions in diverse situations, immediately see the consequences, and reflect on those consequences in terms of specified learning outcomes. Initial results of the analysis of preservice elementary teachers thinking after using the documents based on a framework adapted from the Solo taxonomy are promising.

The purpose of this paper is to report on the emerging results of a project that used an instructional intervention designed to improve middle school students’ informal expectations of variability in a two-dimensional context that was based on the theorized placement of falling raindrops. Specifically, one aim of the project was to compare how students reasoned about variability to make informal inferences both before and after modelling a task physically and then via computer simulation. A simultaneous goal was to have students pursue their own additional questions, beyond the initial prompts given, that were prompted by an analysis of the data they had gathered.

Goal of this research was to develop an abbreviated and statistically robust instrument to assess autonomic symptoms that provides clinically relevant scores of autonomic symptom severity based on the well-established questionnaires. Data from 405 healthy control subjects seen at the Mayo Clinic Autonomic Disorders Center were collected. Length of the questionnaire was reduced from a total of 169 to 31 questions using exploratory factor analysis. Our new simplified scoring algorithm resulted in higher Cronbach alpha values in all domains. This reduced instrument allowed researchers to focus on clinically meaningful variables. Also, shorter survey instrument was less time consuming and less burdensome for critically ill patients, allowing for capturing accurate responses and limiting missing data. Application of exploratory factor analysis in reduction of dimension reduction in this area of neurology research is novel. This reduced survey instrument is now being used to capture data from various clinical studies around the world.

Driving behavior studies often capture electronic measures at 1-30 Hz for long intervals. It is important to find stochastic models that describe such data, with parameters that can be interpreted and accurately estimated. In this report, we review a family of models that are useful in describing the lateral position of a vehicle in a simulator. These models consist of “projection” and “signed error” pieces, with the latter containing a parameter representing the tendency for drivers to return the vehicles to a central position. We use ad hoc and likelihood-based methods to fit these models, but these all result in biased estimates. Fortunately, in two-group studies, simulations suggest that such biases may offset each other and hence that two-group comparisons may have acceptable accuracy. If we can resolve the bias issue, electronic data from a vehicle might be useful in predicting future errors and crashes.

Whereas statisticians of advanced countries are developing new methodologies to reveal patterns and trends inherent in extremely large data sets, the statistics communities of developing countries are deficient in this regard. In the year 2018, a sample survey was carried out in Pakistan to ascertain the level of awareness regarding big data among academics and practitioners of statistics in the country. The survey revealed that, for many terms related to big data, there did not exist much awareness among the statisticians of the country. This paper extends the 2018 study in terms of coverage, scope and depth of analysis. Results of the extended survey seem to confirm the findings of the previous year indicating that there is a need for multi-pronged strategies to create awareness in the statistical community of Pakistan regarding big data and its role in evidence-based decision-making conducive to development and progress of the country.

In sub-Saharan-Africa (SSA), agriculture land-use supports the livelihoods of the majority of people. Land-use for agricultural-activity is an economic-activity that is highly dependent upon weather and climate that produce food and fibre necessary to sustain human life. Hence, land-use for agriculture is expected to be vulnerable to climate variability. This paper examines this relationship. The paper presented data and generated evidence-based decision making under risk and uncertainty as influenced by climate change and its effects on agricultural land-use/outputs. Farm-level cost-route survey of cross-sectional national-data of 800 respondents was used for analysis. Data were analyzed and presented using the tools of descriptive statistics, trans-logarithms model and multivariate probit model (MVP). Study indicated a strong relationship between efficient use of agricultural-land and adaptive-processes to climate-change. Thus, providing data and analysis that strengthen policy decisions on land-use and climate change. Hence, policies of promoting and motivating sustainable land-use management need to be entrenched.

In the context of developing countries, Egypt as well, ‎substantial importance is ‎‎‎always ‎assigned to textbooks. Besides, many teachers ‎place ‎‎‎priorities on ‎discussing its activities. Therefore, in the case of probability, it is crucial ‎‎to ‎reflect ‎on textbooks' ‎discourse for identifying the possible opportunities given to ‎‎students ‎to enhance ‎their ‎probabilistic reasoning. From this ‎perspective, Onto-Semiotic ‎‎‎Approach‏ ‏‏‎(OSA) combined with Batanero et al. (2016) description of main interpretations of probability has been ‎employed to deduce and categorize the essential entities of ‎‎probability longitudinally across ‎‎grades. As a result, the current content can foster ‎‎students' procedural ‎knowledge of objective probability with little attention to the epistemic side that views probability as a personal degree of belief. Consequently, it's ‎important to revise the ‎‎primary school content of probability through ‎implementing some basic probabilistic concepts (e.g., ‎the law of large ‎‎‎number and conditional ‎probability).‎

In times of big data tree-based algorithms are an important method of machine learning which supports decision making, e.g., in medicine, finance, public policy and many more. Trees are a versatile method to represent decision processes that mirror human decision-making more closely than sophisticated traditional statistical methods like multivariate regression or neural networks. We introduce and illustrate the tool ARBOR, a digital learning tool which is a plug-in to the freely available data science education software CODAP. It is designed to critically appreciate and explore the steps of automatically generated decision trees.

Statistical reasoning and the confrontation with first ideas of uncertainty can already be enhanced in primary school. A challenge is how to relate theoretical-combinatorial aspects to empirical frequency aspects, given that fraction concepts are usually not available at primary school. In the frame of a Design Based Research approach we have designed and realized a teaching sequence consisting of seven lessons to develop statistical reasoning about uncertainty of grade 4 students (age 10-11). To supervise their learning processes we collected data on different levels: (a) written pre/post-tests, (b) working notes after each lesson and (c) interviews after the teaching unit. In this paper we will mainly present the design of a teaching unit and first results from the analysis of pre- and posttests.

I conducted experimental lessons that students analyze the data, and make the presentation. This lessons aimed at students’ distinct between “Conclusion from data analysis” and “Judgement of knowledges and experience” in lessons of Mathematics I. I have developed portfolios and make a rubric for criterion, measuring that above objects. I identified growing the attitude of students’ ”construction from data”. It is also identified that students cannot interpret with numeric basis. And growing student fosters attitude interpreting data with evidence. Thus, focusing on criterion of AC phase in actual lessons in PPDAC cycles is better way to achieve above objectives.

Retirement Villages (RVs) are a common form of housing for older people in Australia. RV contracts are very complex. RV residency terminates on death or ill health. At Macquarie we developed a free online RV financial calculator. This is designed to help consumers with understanding the contracts, comparison shopping, and avoiding costly mistakes. It takes account of longevity / health and financial risks. It converts the complex fee structure to a comparison rent payable monthly over the consumers expected healthy lifespan. RVs are much costlier than most consumers expect. The cost varies by gender and increases with age. This tool uses actuarial modelling utilising publicly available data on mortality and disability. The contracts have much in common with insurance policies. This is the first RV calculator available in Australia. The underlying actuarial model it is very original and the calculator can handle the vast majority of contract designs.

Deep Learning Analytics uses predictive models that provide actionable information. It is a multidisciplinary approach based on data processing, AI technology-learning enhancement, educational data mining, and visualization. The problem is that embracing DLA(Deep Learning Analytics) in evaluating data in higher education diverts educators’ attention from clearly identifying methods, benefits, and challenges of using DLA in higher education. Predictive models including random forest (RF), support vector machines (SVM), logistic regression (logistic), and Deep Learning were trained and their performances compared. The predicted value of “source of sustainability” and selected input variables were utilized to predict the drop out of learner. Expected significant outcomes and impact is that using DLA we can find the optimal learning management model for supporting services for instructors significantly impact the quality of statistics education and for learners is necessary to support announcements from instructors, for providing appropriate learning environments.

Fast-and-frugal trees for classification/decision are at the intersection of three families of models: lexicographic, linear and tree-based. We briefly examine the classification performance of simple models when making inferences out of sample, in 11 medical data sets in terms of Receiver Operating Characteristics diagrams and predictive accuracy. The heuristic approaches, Naïve Bayes and fast-and-frugal trees, outperform models that are normatively optimal when fitting data. The success of fast-and-frugal trees lies in their ecological rationality: their construction exploits the structure of information in the data sets. The tool ARBOR, a digital learning tool, which is a plug-in to the freely available data-science education software CODAP can be used for constructing and interpreting fast-and-frugal classification and decision trees. This paper is an abridged version of work by Woike, Hoffrage & Martignon on the integration of classification and decision models in a common framework (Woike, Hoffrage & Martignon, 2017).

Hypothesis testing has come under scrutiny, and attack, because of the way it is being misused. Much of the misuse seems to stem from a fundamental lack of understanding of some key principles within the methodology. In particular, losing sight of the fact that there are two possible wrong decisions. Where p-values are used, and computed by software, it is very difficult to maintain the perspective of the test as trying to identify shifts in parameters – because ‘significance’ has been viewed as the holy grail. The foundations for understanding hypothesis testing are undermined in some curricula where key aspects, such as the existence of two potential errors in the decision, are omitted. This paper will develop a pedagogical basis for teaching the logical foundations of hypothesis testing, and will provide links to electronic resources to support this approach.

A longitudinal study on the evolution of the mental levels of reasoning on decision-making situationally-provoked by a game of chance task to minimize the risk to lose is presented. The task was firstly implemented to 48 Spanish Secondary-school students (age 12). Four years later, it was implemented again to 28 of these 48 students (age 15). A retrospective analysis was performed to identify the stochastic objects involved in the ways of thinking that helped to make decisions and students’ mental levels of reasoning. Students mental levels of reasoning on decision-making evolved from uni-structural responses based on personal preferences to initiate the evolution to extended abstract responses. Three main obstacles constricted a further evolution: the deterministic nature given to the random generator, the lack of proportional reasoning and the ignorance of the relationship between the classical a priori and the frequentist model of probability.

This paper is an advocacy for the adoption of Educational Data Analytics (EDA) as new technologies in evidence-based decision management in Nigerian school system. The paper reviewed the EDA approach not in isolation of other useful theories such as the theory of the application and data analysis in educational institution, the theory of action and organizational support and the process for collecting and analyzing data. The justification for the focus on EDA is that it is now a global trend and that the approaches before now have been passive in the country leading to more problems in the Nigerian education sector. It is also important to note from the paper that one challenge EDA has witnessed in Nigeria educational system is high level data illiteracy by educators hence the need for training and retraining of educators on the use of educational data.

this paper employs data from the Program for International Student Assessment (PISA) 2012 study on mathematics performance in Australian secondary schools to determine the effect of mathematics anxiety on mathematics performance among secondary students. Data of school and student specific factors that are relevant to the Australian educational context are extracted from the PISA 2012 study. These data are used to measure the influence of these factors, as well as mathematics anxiety, on students' mathematics performance. Potential predictive factors are also used in the assessment including gender, socio-economic status (SES) and mathematics anxiety. Findings support the existence of an inverse relationship between mathematics performance and mathematics anxiety whereby the influence of mathematics anxiety varies based on students’ gender and SES.

Unlike mathematics education, statistics education has given little attention to students’ beliefs. In comparison it is possible that statistics may open up another domain-specific horizon of possible beliefs. However, there is no explicit theory about beliefs on statistics. In order to gain insight into students' beliefs about statistics, focus groups of advanced social science students were conducted. The focus groups were analyzed by content analysis and then partly by hermeneutics in order to identify types of beliefs. As a result well-known belief systems from mathematics can also be found in statistics. There are students who view statistics as a system of terms and rules, and there are also students who understand statistics dynamically. The last group can be subdivided into those that extract information out of data and those that want to check theory using data. A fourth group sees statistics as a form of systematic description of reality.

Bringing real problems into curriculum is not an easy job but assessing such student work is even harder, especially when universities are run like businesses where academics are given only limited hours to assess student work. The literature documents that authentic problems are more engaging for students and thus there is no argument against having such problems in the curriculum. However, the literature also shows that assessment of such student work is costly in terms of time. Our experience show that assessments of real projects can be completed more efficiently by utilising marking rubrics. Students seem satisfied with the mentoring and guidance they receive, and with the feedback they get for their projects. Given that such marking rubrics have not been widely developed or used in statistics education generally, nor specifically for real projects. We believe that our examples will be helpful for statistics educators generally.

How many of social sciences students passing introductory statistics courses develop the expected skills to make a meaningful use of statistics? Our diagnosis suggests that an important part of them achieve this through memorization and repetition. This communication reports the in progress effort to improve the quality of the evaluation of an introductory statistics course in Psychology degree, National University of Córdoba (Argentina). There is a specific demand on the qualifications required of students who pass the subject, which combines with a significant volume of students, so it is necessary to ensure the validity of the evaluations and the automation of their administration and correction. The work consists of the construction of examination items classified according to three criteria: elementary thematic unit it evaluates, cognitive level and degree of difficulty, so that precision exams can be built. The proposal is applicable to classroom or on line courses.

When students encounter categorical data, lessons often focus on computing probabilities from two-way tables. These computations may involve simple, joint, and conditional probabilities, and the calculation of relative risk. However, little attention has been given to the questions posed. The purpose of this paper is to explore the questions that undergraduate students pose of categorical data, and their reasoning with a variety of representations of categorical data. Results from a small pilot study suggest that when the questions posed involved making comparisons, students were often confused as to whether they should compare proportions between conditions, or compare proportions within a condition.

Statistical skills are deemed important for psychology students as a prerequisite to learn psychometrics. Thus, the aim of the current study was to identify the learning approach that is more likely to result in better retention of statistics prerequisites to learn psychometrics, and to highlight the individual characteristics of students who adopt it. Data were collected from a sample of students enrolled in a psychometrics course and who had previously passed a statistics exam. At the beginning of the course, several scales were administered to measure statistics self-confidence and attitudes, learning approaches, learning conceptions and teaching preferences, and statistics knowledge. Results showed that knowledge was positively associated to a deep approach to learn, and several individual differences were observed between students who decided to use vs not to use this approach. These findings contribute to current state of knowledge on statistics education and they suggest areas of intervention.

Higher education institutions globally face a continuous expansion of their enrolment in which learner success constitutes a major challenge. Therefore, there is growing interest in the analysis of data linked to student learning engagement. Indeed, large amounts of learning-related student data are currently not being fully exploited, while their aggregation and quantitative analysis would definitely be elements valuable to support teachers and students, to optimize students’ learning experience. In this global context, we have applied, in a public university without any academic filter for enrolment, such analysis to virtually tutor first-year undergraduate students in a statistics course. By supporting them in the form of voluntary online self-assessing tests, we examined what were the personal profiles of the students who were using available tests and how they exploited this help. Finally, using econometric models we tried to determine if there was a link between student success and the use of this help.

What recourse exists when the teaching of Statistics by visual means is not part of the equation? This case study of a team from the University of Coimbra – one teacher, two members of the Media Production Center and a visually impaired student – took up the challenge of teaching a full semester in Statistics in such a way that all students would experience and comprehend fundamental concepts ranging from summarizing data numerically and graphically to making decisions in Statistics through understanding centrality, dispersion and hypothesis testing. A very pragmatic non-visual orientation required class planning to emphasize on contents received aurally or read in Braille, with activities and challenges designed to engage all students in a variety of formats. Materials used in class will be presented in which Statistical Thinking is explored using visual, aural and tactile senses. Future research opportunities will be discussed.

The qualitative case-study reported in this paper is framed by the ‘big’ ideas in introductory statistics, which emphasize statistical thinking, reasoning and literacy. We analysed secondary qualitative data from prospective secondary mathematics teachers’ statements, based on six themes in statistics: the influence of secondary-school experience and the teacher; difficult topics; avoidance of statistics; lack of coverage; teacher knowledge; reasons associated with the difficulty of statistics; and the contributions associated with school practice. We used five elements of the foundations of statistical thinking to enable us to analyse the prospective teachers’ statements from each of the six themes. Data analysis suggests that representing data to enable a clear understanding of statistical concepts emerged more frequently in the pre-service teachers’ statements than the other five foundation elements. The consideration of variability emerged the least number of times. Implications for statistics education are discussed.

To improve the application of statistics within the scientific process, we have developed a course for senior undergraduates in Statistics and early graduate students from any science discipline. The course follows the story of how we use data to understand the world, leveraging simulation-based approaches to perform customized analyses and evaluate the behavior of statistical procedures. It serves as a bridge between learning and applying statistical tools, focusing on statistical thinking within an expanded domain of statistics that includes the beginning of the scientific process, e.g. asking a clear question and tying it correctly to analytical methods, and the end of the scientific process, e.g. communicating results to the general public. We share successes and challenges of our course as well as some examples of in-class activities.

Teachers with positive attitude towards statistics and good statistical knowledge are important in preparing students to learn statistics effectively. This study explores 49 mathematics teachers’ attitude and knowledge in teaching statistics at selected secondary schools in Selangor. Forty-nine mathematics teachers from the urban and rural secondary schools were surveyed using the attitude toward statistics questionnaire and statistics assessment form. Data collected was subjected to an analysis based on the Rasch measurement model. Wright map shows that slightly over 50% of the teachers with positive attitude towards statistics perceived well in their knowledge in teaching certain statistical concepts. Differential item functioning indicates no significance difference in the attitude of teachers between rural and urban schools. Teachers with less than 10 years experience have better knowledge in graph representation and measures of central tendency while teachers with more than 10 years have better knowledge in measures of dispersion and probability.

The current study surveyed Iranian K-12 math and statistics teachers in terms of their assessment literacy (AL). To this end, 135 math teachers were invited to answer a test of assessment literacy, comprising of 31 items on the seven dimensions of assessment literacy expressed in the Standards. The test was tailored to the contingencies of the Iranian education system and was translated into Persian. Consistent with findings from similar studies on teachers' assessment literacy in other educational system, results indicated that Iranian math teachers do not enjoy a sufficient knowledge base in assessment (x̅ = 12, sd= 3.38). Possible causes for math teachers' poor AL and implications the results hold for teacher education are further discussed in the remaining of this paper. The paper closes with a few suggestions for enhancing assessment literacy among math teachers.

Outcome-based education requires statistics education in the 21st century to be structured holistically by allowing the students to work with real life data along with visualization, computation, and learning outputs. The new K-12 curriculum resulted to a chain reaction in the Statistics undergraduate program by including alternative statistical frameworks such as Bayesian statistics. This paper focuses on project-based learning approach on designing learning outputs for undergraduate and graduate students in Bayesian statistics. The stages of project-based learning in completing these Bayesian learning outputs have helped in building the capacities of the students to understand the essential concepts in Bayesian inference and do computations using software. Some of the learning outputs are cited. Insights are helpful in making the syllabus of Bayesian Statistics and Inference of undergraduate Statistics program. Project based learning incapacitates the students to do Statistics research in an organized manner and make decisions based on data

Teaching sampling distribution of the means (SDM) using simulation has the potential to mislead students who might falsely believe that the mean of SDM will more closely approximate the population mean (μ) as the sample size (n) increases. A teaching experiment was conducted involving two Introductory Statistics classes. Both classes were taught the concept of SDM using simulation but the central limit theorem (CLT) was only stated in one class. A questionnaire on assessing students’ thinking and possible misunderstanding about the CLT and sampling distribution was administered before and after the experiment to both groups. A statistical analysis comparing both groups were conducted. It was found that the instructions affect students’ understanding of SDM, however, there is no significant difference in students’ understanding of sampling distribution for both classes after the instructions.

A “Statistics Problem-Solving” course has been designed for the African Institute for Mathematical Sciences (AIMS) and given to 47 students on a Mathematical Sciences MSc in Cameroon. The course exposed students to problems in statistics ranging from design, collection, manipulation and organisation of data through to analysis and reporting through games and simulated and real data. Students also worked in groups to explore and report on a specific problem. These included climate for agriculture, procurement for corruption, a poverty survey and 5 other topics. The students’ evaluations confirmed that the course was an “eye opener” with some students stating a new-found interest despite no previous background in statistics. For others, it was totally different to their past statistics courses. This paper presents this statistics problem-solving course, which was designed to excite and engage students through experiential learning.

Students frequently exhibit randomness misconceptions due to a multitude of reasons. The purpose of this paper is to explore whether auditory cues corresponding to a sequence of simulated events challenge students’ intuitions of random processes. Results from this study indicate that randomness misconceptions were exhibited by the participants and, through running a simulation, their incorrect perceptions were then explored. From this, it is proposed that further research could investigate whether auditory cues are beneficial in challenging students’ randomness intuitions in a classroom setting. The way in which the tool highlighted the students expectations of waiting times generated from a uniform distribution, and constant waiting times was unexpected outcome of this small pilot study and is yet to be fully explored.

Statistics curricula world-wide are undergoing a period of review and reflection. Curriculum change requires an articulation of ambitions and purposes, exemplified by curriculum materials and assessment tasks. Whatever the curriculum specification, it is high-stakes assessments that drive the enacted curriculum. The GAISE Report sets out a broad vision for statistics, and provides an extensive collection of illustrative assessment tasks. We have developed a comprehensive analytical framework which encompasses rather more radical views of what the statistics curriculum should comprise. We classify the GAISE tasks, and identify strengths and areas which will benefit from further development. We provide examples of tasks designed to assess emerging goals for statistics education, and to extend GAISE. Conceptual frameworks are essential for curriculum reform; they offer a focus for discussing curriculum ambition. If new educational goals cannot be exemplified by assessment tasks, then reform efforts will fail.

This paper contributes further insight into the new paradigm of ‘questioning the world’ and study and research paths and proposes that students work with important and meaningful ‘big’ questions. The design of the project proposes ways for teachers to engage students in statistical investigations. In the lessons, the 6 grade students develop critical stance about data, but, in the statistical investigations, it transpires that the questions posed by the students and teacher are insufficient to make the students explain or justify their choices of statistical methods. The students and teacher do not appreciate the potential for statistical inquiry in their own questions, and the students do not question or justify the choice of statistical descriptors to calculate or connect these choices with their answers to the ‘big’ question.

In the last few years, students of Quantitative Analysis of the Faculty of Political Sciences , have had many difficulties in issues related to the application of statistical concepts, which led to lack of motivation and withdrawal from the subject. Considering this problem, in 2018, a new methodology was implemented to evaluate students with the objective of understanding the importance of statistics in the work practice of their career and its applications. This new assessment achieves an improvement in academic performance by having students research and analyze a database, work in groups, and prepare an oral presentation. This new appraisal replaces the written test, which used to generate high levels of anxiety, and allows students to acquire other useful knowledge, such as managing databases, working in groups, using software, etc. Students' status is analyzed through the instrument used to measure the anxiety level GTAI-A.

Initial education is considered as a stage of construction and a transmitter of culture that will – through a variety of games – enable children to explore the world beyond the learning that develops at the affective core. According to Garfield (1995), Garfield and Ahlgren (1998), Kapadia and Borovcnik (1991), and Shaughnessy (1992), teaching statistics should start as early as possible. However, in Argentina, teachers at initial level lack sufficient statistical knowledge. We have designed a teaching experiment with teachers responding to their lack of knowledge, where all activities are fostered around games. The experiment was carried out in Rosario in 2018 with teachers in year 4 and 5 classes, where they were taught notions of statistics in a way so that their pupils will be able to put into action what they learn. The surprising results indicate that statistics and games can establish a profound relationship in the children’s minds.

Teaching of statistics should focus more on practical data science, with a special emphasis on data wrangling: Preparing the data for the analyses, looking at the data via clever visualizations, and learning the principles and practices of open science and reproducible research. The statistics curriculum should be updated and the term “data science” used as a synonym to statistics. In all possible fields, there is a huge need to have more data scientists. To engage everyone with “open data science” (open data, open science, and data science), we have created a new course, where students from all levels and fields work together and share their ideas with openly available data sets and freely available state-of-the-art software tools, such as RStudio, R Markdown, and GitHub. The new course has been quite successful in engaging extremely heterogeneous groups of students to challenge themselves to a “next level” by learning new skills of open data science.

Today we face a more heterogeneous student population in higher education than before and it should be our main objective in this respect to ensure future-oriented, creative and innovative candidates as researchers in STEM subjects, as statistics, when we today face the era of digitalization and big data. In an ongoing study at the Norwegian University of Life Science we try to adapt to this diversity in the introductory statistical course. In 2016 the course was redesigned as a flipped classroom with cooperative learning activities in class. In 2017 further adaption were made: The students that preferred to work alone, could choose to solve problems individually and out of class. Output variables like exam scores and evaluations have been analyzed in light of the learning preferences of the students. Results show, among other, that the so-called digital and introverted students are over-represented in the group that took the course individually.

Amidst the current debate on the practice of statistical inference, more attention should be directed to the random-sampling assumption. We demonstrate the ubiquity of the assumption in sample surveys, controlled experiments and regression models, and argue that the assumption is crucial to inference. If it fails, as is often the case, not only the P value, but also the confidence interval may not be meaningful. An informal survey of papers that present include inferential data analysis found a low rate of a mention of the issue: 1 in 10 papers in two volumes of Journal of Statistics Education, and 2 in 8 papers in one volume of Statistical Education Research Journal. None of the three papers discuss implications of the failure on the conclusions. We make several recommendations to help the statistics education community improve the practice of inferential data analysis.