Why do I see "Trends" in Randomly Generated Data?

Discussion in 'Data Sets and Feeds' started by Rahula, Feb 21, 2008.

  1. In my case those occurrences are in the hundreds of thousands. I was talking to a group the other day and we estimated we have watched, researched or documented over a half million real-time oscillations, each, in a variety of fractal charts. I'd say that was a sufficient size data set to extract some decently reliable data outcomes.
     
    #321     Mar 10, 2008
  2. Jerry030

    Jerry030

    I don't have time to actually run this for you but I'll guess that somewhere between 200 and 2000 observations would give you some good confidence levels, depending on how you structure what you want to prove and the level of confidence required.

    Similar work in pharmaceutical and consumer testing uses about this number of subjects. A recent consumer testing study on weight loss diets where I did the IT consulting but not the statistics had 460 subjects. This was enough to support a claim of efficacy with the FDA.

    For Phase III trials most pharmaceutical companies want several thousand subjects as they have a large liability when a drug goes bad and they get a class action lawsuit.
     
    #322     Mar 10, 2008
  3. Side bar . . . Jerry, are you located in Nevada? Nothing more specific asked.
     
    #323     Mar 10, 2008
  4. Jerry030

    Jerry030

    I would agree that there is randomness. It's a chaotic system which oscillates between totally random and probabilistic behavior. The trick is to distinguish between these prior to making a trade. One method I use is to calculate the correlation coefficient of the predicted behavior of market in relation to its actual behavior at each bar and only trade when the first order derivative of this coefficient is increasing. This of course has a lag based on the number of bars into the future that I’m trying to predict.

    I've never worked with fractals in the market, but use Wavelets.
    Do you have any links to sites that would give some information on fractals in the market?

    People cling to old methods for a number of reasons:

    1) Fear of the unknown

    2) Internal insecurity manifesting as an imprudent dependence on what they already know and an aversion to something new which they may have to learn.

    3) Viewing themselves as an expert in a field and feeling they have to maintain that status by keeping different ideas at bay.

    4) Having failed in an area and wanting others to fail as well...misery needs company.

    Jerry030
     
    #324     Mar 10, 2008
  5. Jerry030

    Jerry030

    No, Chicago a bit west of the CBOT and CME.

    The story is from Wired Magazine.
     
    #325     Mar 10, 2008
  6. MAESTRO

    MAESTRO

    FYI
     
    #326     Mar 10, 2008
  7. Jerry030

    Jerry030

    ProfitLogic....check your ET account for a private message.
     
    #327     Mar 10, 2008
  8. Wavelets and price fractals are interchangeable as I use them.

    Most references to "price fractals" are in relation to Elliott Wave, Gann, Fibonacci, cycles and momentum indicators. I personally do not follow those lines of analysis as I find them too inconsistent.

    I have found that pure price is perfect and then viewing that pure price movement in fractals or, as you are familiar with, wavelets, gives a trader a reasonable and consistent view of trading opportunites inside the larger or longer term moves of price.
     
    #328     Mar 10, 2008
  9. Jerry030

    Jerry030


    Try this one on for size...

    http://citeseer.ist.psu.edu/cache/p...zSzbibliographyzSzlo_mck_88.pdf/lo88stock.pdf

    Stock Market Prices Do Not
    Follow Random Walks:
    Evidence from a Simple
    S p e c i f i c a t i o n T e s t

    Andrew W. Lo
    A. Craig MacKinlay
    University of Pennsylvania

    In this article we test the random walk hypothesis
    for weekly stock market returns by comparing variance
    estimators derived from data sampled at different
    frequencies. The random walk model is
    strongly rejected for the entire sample period (1962-
    1985) and for all subperiod for a variety of aggregate
    returns indexes and size-sorted portofolios........
     
    #329     Mar 10, 2008
  10. Jerry030

    Jerry030

    I agree 100% all you need is OHLC as input and then trasnsform that into predictive elements.
     
    #330     Mar 10, 2008