Directed and undirected networks have been used to describe all sorts of phenomena. There are several approaches to modeling network graphs. Some take a model based approach and others appeal to directed graph methods. An excellent jumping off point is Statnet.
Software development efforts at Tufts School of Medicine is addressing Meta Analysis needs with a free PC based Windows program known as MetaAnalyst. Their program provides several models for discrete, continuous and diagnostic data. MetaAnalyst also integrates BUGS for Bayesian meta-analyses options.
This new toolbox product was recently added and made a default part of the Matlab suite in the latest version; a good reason to upgrade!
Complex multilevel designs present additional statistical issues when it comes to addressing suitable sample sizes. OD is a software tool that address many of these issues for simple and complex multilevel designs. It seems very useful for planning and reviewing assumptions in existing designs.
Very often a new research design is confronted with the question of how much data do I need…? Power Analysis software is widely available to address many of the standard fixed sample size solutions to this question and its many variants. Use of these tools can be difficult from the researcher’s point of view in the sense of prior information elicitation needed as input. StatXact software has 3 products, East, Compass and SiZ, that should be considered for dealing with many of these tasks. However, recent work on adaptive power designs has added new options that adjust required fixed sample sizes at interim points of implementation so that desired power is targeted. A paper by Metha and Pocock, Adaptive Increase in Sample Size when Interim Results are Promising, presents the theory and case studies illustrating the advantages of Adaptive Designs.
After surveying departmental chairs at the Tufts Medical School and Sackler School of Graduate Biomedical Sciences, UIT has purchased the licensing of the SAS Genetics module as an add-on to the SAS Educational Analytic Suite software bundle that Tufts renews every year. The SAS Genetics module is now available, at no additional cost to users, for installation with every SAS license that has been purchased under Tufts’ annual volume license renewal. The only restriction pertains to the SAS version in use. SAS has only released the Genetics module for SAS version 9.2 32-bit and SAS version 9.2 64-bit. Therefore, use of the new SAS Genetics module may require upgrading to one of these 2 versions of SAS. Please contact your local FSP(IT support person) if you need to purchase SAS or have SAS Genetics added to your current SAS installation.
Systat has been a full feature stats package for some time and has a relationship with Spss. One of it’s sub- products is called SigmaPlot. If you only need access to a wide variety of easy to access graphics and some basic stats functionality you might like this product. Most everyone finds it very user friendly. SigmaPlot is available under Tufts’s network concurrent license. Along the sames lines, if you are looking for a free student version of a statistics package, try Systat’s MyStat. What a deal!
Statistical spatial analysis methods are evolving at a rapid rate as a result of increased active basic research and new journals promoting such methods in applied settings. One such journal is the Journal of Agricultural, Biological, and Environmental Statistics. An article by authors Woodard, Wolpert and O’Connel, Spatial Inference of Nitrate Concentrations in Groundwater, June 2010, illustrates nicely the application of nonparametric bayesian kriging via a spatial moving average model. This approach improves upon traditional kriging methods in many ways. The authors provide an S-Plus software package, spatialLrf, to implement the new algorithm.
Multilevel models are finding widespread use in many social science settings. The traditional settings have been developed in the statistical mixed model setting. Extensions to non normal models are also finding useful applications. These models are typically expressed as a hierarchical model where parameters are expressed by their relevant distributions. Specialized software solutions are needed to deal with the estimation difficulties. The Centre for Multilevel Modelling is a very useful resource for both theory and software and is the home of WLwiN. Of particular note, is the page on software reviews, where many software packages are discussed and sample codes provided. Be sure to check out the discussion and software devoted to multiple imputation in the multilevel model setting! Last but not least, take a look at the work done by a major contributor to multilevel models, Harvey Goldstein, addressing Multilevel Structural Equation Models as implemented in MLwiN.
Like many universities, Tufts has a subscription membership to the ICPSR. Over the years, a typical use case would be for a researcher to download data for local analysis. In this way one would have maximal access to raw data and use of analysis tools of one’s choice. Recently, ICPSR has provided a web front end tool, Survey Documentation Analysis (SDA), for online analysis suitable for routine statistical queries. Many datasets in the archive have been prepared for use with this tool. SDA is available to any Tufts faculty, staff or students. It would seem that use of this interface in support of some class teaching activities would lower the barrier to analysis that a more traditional approach would require.