[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Chi-squared? Random distribution != random process



PureBytes Links

Trading Reference Links

What can one conclude if the distributions are close enough by the chi^2 measure? In particular can we infer something about the veracity of using TA? Frankly, I think the answer is no, at least not always. The distribution and chi^2 does not say anything about the time behavior of a process (should be obvious since the data is treated as a collection, not as a series!).

Consider the following degenerate, contrived, hypothetical data. (Notice there is a use for synthetic data... ) The attached image has two curves - one is a normal distributed random sequence, the next is the same set of numbers sorted in ascending order. The numbers are exactly the same, chi-squared is 0 and, hence, both derive from the statistically same set of randomnumbers.

Clearly, both curves are not a "random walk process". The monotonically increasing sort is clearly dependant on history, the next number is always selected to be greater than all preceding. If I can find a stock with that behavior I'm home free - any thoughts Steve <g>.

Anyway, this helped to clarify my thoughts. Perhaps you will also find it stimulating.

Cheers,

Richard





------=_NextPart_001_0250_01C21852.5BCF91B0
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2716.2200" name=GENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY>
<DIV><FONT face=Arial size=2>I have been thinking about the meaning of 
chi-squared tests (and others).&nbsp; As I understand chi^2, it compares 
distributions, basically looking at a variance of the test vs. the sample 
distribution histograms.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><FONT face=Arial size=2>What can one conclude if the distributionsare 
close enough by the chi^2 measure?&nbsp; In particular can we infer something 
about the veracity of using TA?&nbsp; Frankly, I think the answer is no, at 
least not always.&nbsp; The distribution&nbsp;and chi^2 does not say anything 
about the time behavior of a process (should be obvious since the data is 
treated as a collection, not as a series!).</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><FONT face=Arial size=2>Consider the following degenerate, contrived, 
hypothetical data.&nbsp; (Notice there is a use for synthetic data... )&nbsp; 
The attached image has two curves - one is a normal distributed random sequence, 
the next is the same set of numbers sorted in ascending order.&nbsp; The numbers 
are exactly the same, chi-squared is 0 and, hence, both derive from the 
statistically same set of random numbers.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><FONT face=Arial size=2>Clearly, both curves are not a "random walk 
process".&nbsp; The monotonically increasing sort is clearly dependant on 
history, the next number is always selected to be greater than all 
preceding.&nbsp; If I can find a stock with that behavior I'm home free - any 
thoughts Steve &lt;g&gt;.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><FONT face=Arial size=2>Anyway, this helped to clarify my thoughts.&nbsp; 
Perhaps you will also find it stimulating.</FONT></DIV><FONT face=Arial size=2>
<DIV><BR>Cheers,</DIV>
<DIV>&nbsp;</DIV>
<DIV>Richard</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><IMG alt="" hspace=0 src="cid:024e01c2187c$44a105d0$6400a8c0@xxxxxxxxxx"; 
align=baseline border=0></DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV>
<DIV><FONT face=Arial size=2></FONT>&nbsp;</DIV></BODY></HTML>

------=_NextPart_001_0250_01C21852.5BCF91B0--

Attachment: Description: ""