One of the IVZ's most important fields of work is collecting health and health care related data. IVZ then processes these data for further analyses, compiles standard reports and yearbooks and reports to institutions Slovenia is required to report to. Data analysis also underlies the foundation for the majority of the Institute's research activities.
We will outline health and health care related data sources emphasising the data managed by the IVZ. We will review the legal basis for data collecting and give a brief outline on the history of the creation of IVZ-managed data collections. We will discuss the purpose and contents of the key data collections and explains some of the limitations that apply to their usage. Selected examples from the data analyses conducted by the Institute will be presented.
You will also learn what projects - in the field of data collections - the Institute is currently active in and what procedures apply to using the data managed by the IVZ.
Slides: PDF
In statistical analyses - and also in web survey data collection - we mostly use software in English language. With that we also avoid the problem of translation English terms into Slovenian language. In the lecture, a web survey software tool developed within Centre for methodology and informatics, Faculty of Social Science, Universtiy of Ljubljana, will be presented - the "1KA" (EnKlikAnketa) or 1CS ("OneClickSurvey"). Within this context, basic features of 1CS tool will be outlined questionnaire design, data collection, data management and on-line statistical analysis. The problems of translating key terms in this professional field from English to Slovenian will be also discussed (e.g. "grid", "radio button", "checkbox", "drop-down", ...).
Slides:PPT
Photos: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14.
The number of questions (i.e. statistical hypotheses) in biological experiments is constantly increasing and the ability to understand the results at a glance it is often beyond human capacity. Then subdividing the problem in different aspects (or clusters) becomes a natural solution in order to get understandability of the problem.
In many cases the problem can be naturally subdivided into domains in a hierarchical structure. As an example, consider Genome-Wide Association Studies (GWAS) where it becomes natural to subdivide the whole genome in chromosomes, and then in genes (or regions) and finally in single SNP.
This is also the case in most multiple endpoint trials and in self-reported psychological and sociological questionnaires where many aspects (endpoints or items) contribute to describing a more complex concept. When, however, such structure is not given, it is also possible to define it through hierarchical clustering.
A recent work of Meinshausen (2008) makes hierarchical testing of hypotheses possible while controlling the Familywise Error Rate. The method has comparable power to the Bonferroni-Holm procedure for the final levels of the hierarchy, but increases the power on intermediate levels (such as the gene or chromosome levels on the GWAS example).
After a short introduction to the multiple testing problems, we will show how Meinshausens method fulfills the conditions of the Sequential Rejection Principle (Goeman and Solari, 2008). Starting from this result, we show 1) how to extend the Meinshausen procedure to non-binary and/or unbalanced hierarchies and 2) how to uniformly increase its power. We finally prove that the method also works when parent hypotheses are not defined as the intersection of descendant hypotheses.
Two real microarray data example are used as leading example and are discussed.
N. Meinshausen (2008) Hierarchical testing of variable importance. Biometrika 95(2), 265-278
J.J. Goeman and A. Solari (2008). The sequential rejection principle of familywise error control. Submitted
Slides:PDF
Photos: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10.
Odds ratio (OR) is a statistic often encountered in medical literature. Most readers perceive it as relative risk (RR), although they usually don't know why that would make sense. Since this perception is usually correct, there is nothing wrong with it. Still it is probably useful to be occasionally reminded about the connection between the odds ratio and the relative risk, to avoid situations when such equating is wrong.
Another statistic which is also understood as an estimator of the relative risk is the hazard ratio (HR). It is always reported when we use the Cox proportional hazards model to analyze survival data. Under proportional hazards a 'natural' argument is this: if at any time probability of dying in one group is k times the probability in the other group, then the relative risk must be k, regardless of where we are in time. Well, we shall see if this really is so.
Slides:PDF
Confidence intervals are one of the most important concepts in statistics and are standard tool for presentation of estimated parameter. Unfortunately, concept of confidence intervals seem to be confusing and is mostly misunderstood by students and users.
Confidence intervals are tightly connected with concepts of standard error and sampling distribution. Sampling distribution is the first step from descriptive statistics to inference and generalization of sample statistics to population parameters. The crucial question in this extension is: how close to the true parameter value is an estimate from a given sample. The sampling distribution properties, such as standard error, provide information about the spread of estimates. The precision of estimate and possible location of true population parameter can be described with standard error and derived confidence interval.
The concepts of precision of estimate and confidence intervals will be demonstrated with the use of dynamic graphics and computer simulations. Simulations will also reveal the influence of sample size on shape and position of sampling distribution, shape of sampling distribution for different estimators as well as the properties and true meaning of confidence intervals.
Slides:PDF
Photos: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Background: A new microarray platform (GeneChip Exon 1.0 ST) has recently been developed by Affymetrix. This microarray platform changes the conventional view of transcript analysis since it allows the evaluation of the expression level of a transcript by querying each exon component. The Exon 1.0 ST platform does however raise some issues regarding the approaches to be used in identifying genome-wide alternative splicing events (ASEs). In this study an exon-level data analysis workflow is dissected in order to detect limit and strength of each step, thus modifying the overall workflow and thereby optimizing the detection of ASEs.
Results: This study was carried out using a semi-synthetic exon-skipping benchmark experiment embedding a total of 268 exon skipping events. Our results point out that summarization methods (RMA, PLIER) do not affect the efficacy of statistical tools in detecting ASEs. However, data prefiltering is mandatory if the detected number of false ASEs are to be reduced. MiDAS and Rank Product methods efficiently detect true ASEs but they suffer from the lack of multiple test error correction. The intersection of MiDAS and Rank Product results efficiently moderates the detection of false ASEs.
Conclusion: To optimize the detection of ASEs we propose the following workflow: i) data prefiltering, ii) statistical selection of ASEs using both MiDAS and Rank Product, iii) intersection of results derived from the two statistical analyses in order to moderate family-wise errors (FWER).
Slides:PDF
Photos: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
The seminar will discuss some issues related to the analysis of life insurance contracts. These contracts depend on demographic factors, financial factors, or both. After reviewing the classical approaches to the valuation of securities in efficient markets and the underlying economic assumptions, the seminar will focus on how these ideas can be exploited for the pricing of life insurance claims. Some attention will be devoted to the joint modelling of financial and demographic uncertainty.
The lecture is aimed at getting familiar with the focus group method, which is one of the qualitative methods for collection, analysis and interpretation of data. First, the method will be presented with its characteristics, advantages and limitations. In the second part, the practical aspects of designing and implementing focus groups, and analysing and interpreting the gathered experiential material, will be presented. The practical part of the lecture will supply samples of materials required in different steps of carrying out a focus group.
additional literature
Slides: PDF
Photos: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
A large class of clustering problems can be formulated as an optimizational problem in which the best clustering is searched among all feasible clustering according to a selected criterion function. This clustering approach can be applied to a variety of very interesting clustering problems, as it is possible to adapt it to a concrete clustering problem by an appropriate specification of the criterion function and/or by the definition of the set of feasible clusterings. Both, the blockmodeling problem (clustering of the relational data) and the clustering with relational constraint problem (clustering of the attribute and relational data) can be very successfully treated by this approach. It also opens many new developments in these areas.
Slides: PDF
Photos: 0, 1, 2, 3, 4, 5, 6, 7, 8