Archive for March, 2009

Developmental errors in aggregate trending models

Thursday, March 26th, 2009

As discussed previously, while model overlays are typically most useful for analyzing the growth pattern and potential strategies of specific businesses or within more localized market segments, aggregate trending models tend to be most useful for identifying potential model anomolies within related market sectors and predicting the likely future existence of breakthroughs that could lead to rapid acquisition of market share.

The downside is it is still nearly impossible to predict WHAT those breakthroughs will be other than by using relatively new analyis techniques.

And that’s still a big problem. In order to undertake proper development of models to analyze aggregate trending models, multiple output scenarios have to be posited and “assumed” to some degree, but other than comparing the model results to real world results it is for the most part very difficult to accelerate the detection of predictive differences off the posited models.

Posted in Known Issues | No Comments »

Model-independent overlays - the ultimate goal?

Sunday, March 1st, 2009

OK. We’re going to “geek out” on you a bit here. There’s been a lot of buzz recently about model-independent overlays when it comes to collecting and organizing data for feeds into aggregate trending models. As in any good system the separation of data from the mechanism for manipulating the data is key although in certain instances there is cohesion that can’t be completely avoided.

However, there can be sets of overlays that can be tied to specific instances of data organization and theoretically the data modeling can be optimized in such a way to maximize the number of potential overlays within certain overlay constraints. However, once again, the data and the overlays can’t be totally decoupled. That’s not hard to grasp from a “common sense” perspective, but as it turns out it’s quite difficult to prove conclusively.

Once again the time-factor is non-trivial in relation to the coupling/decoupling of data/overlay connections (also known as “n-realms”). Of course when dealing with any type of data collection that is derived from or has a bearing on specific markets and sub-markets naturally the element of time will be an important factor. But what’s surprising is the WAY in which time is a factor when manipulating data/overlay connections.

Posted in Musings & Miscellaneous | No Comments »