 |
| 2014.03.07 DJIA Long Term Cycle Chart |
 |
| 2014.03.07 DJIA Long Term Cycle Chart Z-score |
I developed vs 3.00 of the DJIA cycle model after VS 2.0 failed to converge. Unfortunately, I don't have as much confidence in this model as I do in the other cycle models because even though it passes the predictive testing on historical data, so did the other two DJIA models that failed going forward. Given the model prediction above, it should be apparent by mid-Summer 2014 if the model is on track. Assuming of course it does not "blow-up" before then! I have included the Z-score for the model, which answers the question: How far out of line is the past prediction compared to historical data? As you can see, already we are at an extreme similar to 2008.75.
3 comments:
Thanks for sharing your updated DJIA model, which seems to agree with Elliott Wave predictions of a strong correction, after which new highs are expected.
Your model also concurs with regular technical analysis using trendlines and indicators (like SmarTrend). Cycle analysts that use Hurst signals also anticipate a top, not to mention fundamental analysts that keep pointing to stacks of sentiment and economic measures.
The next several weeks will be most interesting indeed, especially if the FOMC meeting next week yields an unexpectedly large taper.
Kind regards,
Mr Anon
Oops - I forgot to ask what changes you made to your previous model, at least in terms to someone technical, but not a specialist in cycles. It seems that the predicted path has more components or a finer temporal resolution? The double peak compared to the previous single peak suggests that to my inexperienced eyes.
Mr Anon
Thanks for dropping in Mr Anon. I restructured the model to avoid too many parameters, which in model-speak results in an overdetermined model. If the model coughed up a particular cycle, my threshold for accepting it was higher than it was in the previous models. A good pointer to the concept is here.
http://www.mathworks.co.uk/matlabcentral/newsreader/view_thread/307674
Post a Comment