jack you obviously answered to a post of mine. unfortunately, and maybe because of your superior understanding, I do not get your point. maybe you drop another line and specify what you are criticizing? peace
what jack is saying is a basic engineering concept. If you have data that is reported to one decimal place, obviously that is the limit of the granularity. You cannot derive something from that data that purports to go to three decimal places, for example. I'm not commenting on the other posts because they are a bit beyond my knowledge but that is the concept of significant figures.
I see. I agree that differences beyond 0.999 are generally troublesome. But I do not think that this is due to the fact that only two digit numbers are used as an input, since in correlation analysis the number of observations is the key not the analyses data itself. For instance you can make correlations in quantum physics which deal with numbers far behind any the zero, or use number in astronomy far in front of the zero. If you have enough numbers you can build reasonable correlation figures at any digit after the comma with these very different inputs. For whatever reason this thread seems to have ended anyways. peace
There are some really good ways to improve on making money. The statistical stuff is usually not one of them. The interrelationship of things is very important but what is not related, sometimes is more helpful. you will probably find that the more statistuclating that is done the closer one gets to a macro viewpoint. This is precisely driving you away from the micro incedental process of appreciating capital with the highest money velocity possible. There are some really stupendous ways to get out of the box ac ..... is in.
You are incorrect. The important numbers are moved to where they are by bunches of zeros which are merely place holders not things that define accuracy. Further, you are in a place where you cannot see the low value of acrary's input here and elsewhere.
by your strange posts, jack hershey. offense without information. I am not an english native speaker and that might be the reason why I do not get the essence of what your trying to say. If you think I am not in the position to understand "acrary's low value" (quite insulting BTW), then give me a clue of what the real thing is that I am missing. Or put me on your ignore list. peace
Corrolation and dependency are neat topics for making money. I'm sure for you there is substantive content in what you provide and what you read form others than myself. I will try harder in the future to give regard to including substantive content in what I say. It is simply my belief that you cannot get better results than the quality of the subject data. Naturally I choose to suggest that a person who wants to be most successful, should have a focus on the corrolation and or dependency matters that specifically deal with a focussed subset of the whole. The yopics of significant figures and statistical significance is where the basis of my comments may be found. The macro business in dealing with markets, indicators and strategies is not a fruitful nor fertile place for getting the job done. There are several reasons for this. People may do as they wish. However, anyone who is earnest about his efforts has to begin to focus on principles at some point. If a choice is to be made between macro or micro analysis and research to make money, then the macro choice is not viable ever. In any pursuit there are tough things to deal with. I am a person who has always focused on what seems to be tough situations that need to be dealt with. I apparently offend you, insult acrary, and only try to do what I do and it comes out poorly. Naturally as a consequence, you recommend that I either shape up or put you on ignore. MY view of making money in markets and using every skill possible is one where a supreme focus is required as I deal with a near perfect subset of performing entities. I use the major variables of the market and interrelate them on the trading fractals where capital appreciation is optimized among the variables and in order of their significance. Think of this as sculpting a creation by seeing the opportunity and sheaving and sluffing off massive chunks of detris to get to the core of the opportunity and display it to any and all who wish to see what is possible from the raw material available. While this typing is merely run on copy to address you views, get the picture that I am coming from a place where for nearly 50 years I have been creating a model of excellence. I have paid my dues and I know the name of the game. Acrary's stuff and your comments are missing the target in several ways. Attached is a daily analysis chart that will give you and idea of what corrolation and dependence really are. If I wanted to step up to a high level of oppotunity definition with you I would begin at a matrix level.
It seems combining good backtesting results and proper testing of variables for correllation and dependency would be a quite interesting issue.
My current understanding (mostly collected from textbooks) about (commonly discussed) statistical correlation: - The two variables both are basically Random variables; and normally distributed (whether independently or not); - The relationship is usually based on simple Linear; It measures mainly the Strength (rather than the Form) of any linear relationship; - A zero (coefficient of) correlation does not imply no relation exists; instead it indicates the variables are not (i.e. zero degree) linearly related/ correlated; therefore the two variables that have a very-weak/ close-to-zero linear correlation could have an extremely high degree of (say) curvilinear relations; - If the null hypothesis that a zero (coefficient of) correlation can be rejected, it can be concluded that there is a linear relationship (of whatever degree which can be anyhting greater or smaller than zero) between the two variables; - A high degree of correlation merely indicates the association is close to linear; Any perfect correlation occurs only when the points lie exactly on a straight line; - The coefficient can be strongly affected by a few outlying observations; therefore it must be evaluated carefully with caution when outliers appear in the scatter-plot; - A highly strong correlation should not be interpreted as any direct relationship of cause and effect; the relationship can be simply nonsensical; - The scatter-plot is not absolutely reliable by merely visual inspection since the plotting scales can be adjusted and mislead; - Correlation would not be a complete description about the relations of two variables for trading purpose;