Generational Dynamics Forum - New Site

[quote="mww"]
> John - would you say that the U.S. is completely blameless in the
> Syrian civil war? Have we not provided support to those working to
> overthrow the Syrian government in the forms of arms, training,
> intel, and even air strikes? One could say that without our
> support, the war would have likely fizzled out by now.[/quote]

Our support has almost entirely been targeted at ISIS. Turkey and
Saudi Arabia would have supported the rebels with or without US
involvement. The entire Mideast is headed for a major war, no matter
what the US does.

[quote="Jack Edwards"]
> Well that was long.[/quote]

That was a great response, Jack. Thanks

[quote="Jim4881"]
> Just curious but not overly so.... How are your statistical models
> useful if often incorrect – unless they are not meant to actually
> be applied to anything of value or unless you are working through
> a government grant? Unless they are for amusement only, I miss the
> point of the exercise.[/quote]

The Generational Dynamics methodology always works, as long as
you follow what I call "the Chaos Theory rules," one of which
is that you can't predict the actions of one person or a group
of people.

I described some of this stuff in my 2009 paper:

** Generational Dynamics Forecasting Methodology (PDF)
** http://GenerationalDynamics.com/gdgraphics/CSC_Papers_2010_Generational_Dynamics.pdf

Note particularly at the distinction between "short-term forecasting,"
which produces forecasts that are for a small window of time, usually
a few days, but are only 50-60% likely to be correct, versus
"long-term forecasting," which produces forecasts for a large window
of time, usually a few years, but are almost 100% likely to be
correct.

As for your question about whether "short-term forecasting" models
are of any use at all if they aren't guaranteed to be accurate,
here are two such models that I depend on every day:[list]

[*] Weather forecasts are based on models that are rarely accurate for
more than a few hours. A weather forecast that's more than a few days
out is practically useless. Still, I check the weather forecast each
morning to decide whether to wear a raincoat.

[*] I take a bus to work each morning, and I check a mobile app that
predicts when the bus will arrive, so that I'll know when to leave my
home for the bus stop. I've discovered that the predictions are
frequently inaccurate, but the inaccuracy is always on the side of
being five or ten minutes too early, so if I get to the bus stop at
the predicted time, then I usually don't have to wait more than five
or ten minutes.[/list]

So imperfect models are used by everyone all the time.

Jack describes how he uses his scientific models, and how he adjusts
them when necessary. In many ways, the value of a model lies in the
process of how the model is used, as well as in the model itself.

John - would you say that the U.S. is completely blameless in the Syrian civil war? Have we not provided support to those working to overthrow the Syrian government in the forms of arms, training, intel, and even air strikes? One could say that without our support, the war would have likely fizzled out by now.

Jim4881 said: [quote]Just curious but not overly so.... How are your statistical models useful if often incorrect – unless they are not meant to actually be applied to anything of value or unless you are working through a government grant? Unless they are for amusement only, I miss the point of the exercise. My winning Lotto model has been adjusted too many times to mention, but I am only misleading myself since the variables keep resetting.
[/quote]

It's a good question. Most of my models relate to chemical industrial processes. When I build the models they often stay accurate enough to help me predict how to optimize things for days/weeks/months but usually not years. When the model stops predicting accurately I ask myself why? Invariably it comes down to a change in my inputs, an inaccuracy in my measuring devices or some other variable I'm not measuring accurately or at all. Often it just needs a little tweak.

For instance I don't know how to model scale buildup inside the process. I don't know that that phenomena is causing the drift in the model but I suspect it is and I know that after I do a major cleanout, that the model changes. The scale buildup has too many factors influencing it, so I can't accurately predict it yet and don't know of a physical sensor that can reliably measure it.

Also - when my model stops accurately predicting, it may lead me to discovering that something is "wrong" - (ie. something needs to be mechanically fixed). So a model that doesn't predict the variable I want - may help me discover something else I needed to look at. Sometimes it leads me to invent new measurement techniques. I often learn good things from a model gone awry. And - often I never figure out why the model went awry. But - by doing these things, using these models and applying learnings to optimize other things I've been able to save the business I work for plenty of money.

As to why bother if you know it's going to eventually stop working.. or may be inaccurate? What's the other option? Wander around blind? Completely oblivious? I'd rather have a model or multiple models and some healthy skepticism to guide me than nothing. That and I find modeling things fascinating. One of my bigger challenges is that some people in management who are frankly statistically ignorant don't understand this concept and are ready to toss out the entire model when it's off 2%. People that don't understand how to model think it's this magic thing. It's not and that concept is germane to this whole discussion.

Now what I do is complicated, but not that complicated. It's stuff you could learn how to do with the right kind of college - jr/sr level statistics class - that and a fairly expensive software package. I've had a couple college classes on modeling and done lots of poking around to understand it better, but I by no means consider myself an expert. Just a guy whose done it a lot and learned things along the way.

John's model works with large sets of people and historical occurrences which is much more messy mathematically speaking. His input measurements are also much more complicated and less reliable. How do you measure the general anxiety level in a population, a tribal geographic boundary, the effect different languages have towards creating animosities between cultures, how does population density affect how groups would respond to crisis events (for instance)? Having those inputs and thousands more might improve the model - but they aren't easily obtainable or quantifiable.

Don't be so critical. The generational dynamics model holds up really well, particularly since it's working with such a messy data set. Read what other prognosticators think is going to happen. Compare and contrast. It's ok to be skeptical - any good scientist is. I just think it inappropriate to junk a model that's been remarkably accurate in the past if all it needs is a little tweak. Any model maker that thinks his model is perfect is deluding himself. So John saying he's adjusting the model based on new information only makes it better in my opinion.

Well that was long.
Jack

Guest (Jack) said …
“As part of my profession - I build statistical models. They often end up being incorrect and I adjust my model.”

Just curious but not overly so.... How are your statistical models useful if often incorrect – unless they are not meant to actually be applied to anything of value or unless you are working through a government grant? Unless they are for amusement only, I miss the point of the exercise. My winning Lotto model has been adjusted too many times to mention, but I am only misleading myself since the variables keep resetting.

Same is true for GD since to this point and excluding the Nukes there are, and will be, fewer and fewer individuals within any advanced (1st world?) generation who have actually experienced the horrors of a war due to advanced weapons systems no matter how broadly defined. Still may apply to primitive peoples to a large degree .... but then there is always a madman lurking. Interesting but pointless .. to me admittedly without special knowledge or experience.

[quote="jimw4881"]
> John ... you could have stopped with "Yeah, you're right"
> ....[/quote]

Or I could have summarized the entire response with, "Yeah, you're a
moron," which was my initial temptation.

[quote]How do you explain WW 2? The WW 1 generation was running the show. Why didn't the WW 1 generation stop it? I'm not attacking you, John, but I'm curious.[/quote]

http://generationaldynamics.com/dl/GenerationalDynamicsForecastingAmericasDestiny.pdf

page 44

The entire book is worth a read.

Jack

John ... you could have stopped with "Yeah, you're right" ....

How do you explain WW 2? The WW 1 generation was running the show. Why didn't the WW 1 generation stop it? I'm not attacking you, John, but I'm curious.

jimw4881 said:[quote]Well then, are not all of Generational Dynamics “predictions” at least unreliable? The fudge used to be “time”, how fast or slow would the GD predictions come to pass which allowed for unlimited "wait and see" statements that were made on the assumption that they will prove to be true. While your posts will remain entertaining they are now officially uninformative and, according to you, unreliable in any predictive analytics sense with this admission.[/quote]

George E. P. Box (a statistician) said: "Essentially, all models are wrong, but some are useful."

As part of my profession - I build statistical models. They often end up being incorrect and I adjust my model.

I find John's model of Generational Dynamics extremely useful - so much so that I read it daily and have for years. It helps me make sense of the world. If his model is off - it's not by much. It's a really good model. Him noting that he adjusts his model is a good example of how a scientist should be. It's more accurate than any model I've made and FAR more complex.

I don't understand why people that don't agree with John bother to post stuff like what you just did. It happens all too frequently. If you think it's a waste of time don't you have other things you'd rather be doing?

Jack