Paul, could you be more explicit in what you mean by "aggregate statistical methods"? Thanks. On Thu, Apr 10, 2014 at 2:53 PM, Paul Barrett <[log in to unmask]> wrote: > Elsevier, the publisher of: > > *Grice, J. (2011).* Observation Oriented Modeling: Analysis of cause in > the behavioral sciences. New York: Academic Press. ISBN: 978-012-385194-9, > > has identified OOM as an innovation in data and causal analysis that > should be made more widely available than at present *(before, you could > only use it if you purchased the book).* > > > > To that end, they have encouraged James to set up a specialist OOM website > from which the software may be downloaded, along with a host of video > examples, and publications explaining, promoting, and using OOM as the > analysis tool. > > http://www.idiogrid.com/OOM/ > > > > The latest publication from James is beautifully written, cogent, with > good solid examples drawn from the textbooks and re-analyzed, with some > great teaching suggestions for those methodologists who may wish to impart > knowledge of this approach to analyzing data and developing/testing causal > models. > > *Grice, J. (2014).* Observation Oriented Modeling: Preparing students for > research in the 21st century. *Innovative Teaching*, 3, , 1-27. > > > > I have not used OOM myself for any major analysis because much of what I > do mirrors the logic, philosophy, and purposes of what OOM sets out to > achieve. > > > > *But, that is about to change*. James and I are about to undertake a > series of I/O domain analyses, showing how to powerfully answer > straightforward validity questions without the statistical tomfoolery that > epitomises current approaches. > > > > Having read the Innovative Teaching article .. I can now see how I would > have approached the t-test example contained therein myself, again without > ever having gone near a t-test or aggregate statistic. The reason why is a > consequence of the kind of hypothesis statement you seek to test, and how > you address the analysis outcome in terms of the claims you wish to make > about the 'effects' or 'model' you have fitted to your data. > > > > I want to explain this in more depth in an article with James, where I > approach a typical published dataset which uses aggregate statistical > methods, compare these with OOM analyses, compare these with how I would > have sought to test the same hypotheses using my own approach *(and how > and why)*, then we draw the linkages between all three kinds of analysis, > showing why aggregate statistical methods can never say much of any > scientific interest. > > > > What you will also see is that I have to computationally build solutions > for almost every problem I meet, drawing upon simulation, randomization, > pattern-matching algorithms, and whatever else I can scavenge from the > analysis-methodology toolbox; inventing stuff along the way if needs be. > Which is exactly why OOM is so powerful and useful. It is an integrated > system for handling a huge variety of analysis designs, without the need to > program and construct afresh almost every analysis. *But to use it you > have to think as a scientist, not as a statistician*. > > > > The Observation Oriented focus is more demanding *scientifically* on > investigators because error-observations are made explicit, and these > require an explanation if your theory-driven statements, or even mild > hypotheses, are to retain their validity. Contrast this with those using > ANOVA, SEM, latent variable modeling, HLM, moderation/mediation methodology > etc. Explanatory model-accuracy is not even addressed, let alone examined > at the level of the observations themselves. > > > > But explanatory model accuracy goes hand-in-hand with any OOM analysis. > That is a profound consequence for both scientific and pragmatic real-world > applications. > > > > Regards .. Paul > > > > *Chief Research Scientist* > > *Cognadev.com* > > > *__________________________________________________________________________________* > > *W*: www.cognadev.com <http://www.pbarrett.net/> > > *W*: www.pbarrett.net > > *E*: [log in to unmask] > > *M*: +64-(0)21-415625 > > >