Dependency statistics in quotes (information theory, correlation and other feature selection methods) - page 70

 

Moving on to Klous

Chart

The graph is different in relation to the opener

Descriptive statistics

And the stats are different. That's amazing. Always thought open and clowes were the same values.

ACF

Date: 10/14/12 Time: 13:53

Sample: 1 100

Included observations: 99

Autocorrelation Partial Correlation AC PAC Q-Stat Prob

.|* |00 .|* 1 0.077 0.077 0.6031 0.437

.|. | .|. |. 2 -0.038 -0.044 0.7502 0.687

.|. | .|. | 3 -0.038 -0.032 0.9001 0.825

*|. | *|. | 4 -0.181 -0.178 4.3327 0.363

.|. | .|. | 5 -0.013 0.012 4.3511 0.500

.|. | .|. | 6 -0.017 -0.034 4.3810 0.625

*|. | *|. | 7 -0.127 -0.139 6.1421 0.523

.|. | .|. | 8 0.048 0.035 6.3987 0.603

.|* | .|. 9 0.086 0.069 7.2140 0.615

.|. | .|. | 10 0.011 -0.015 7.2283 0.704

*|. | *|. | 11 -0.089 -0.136 8.1289 0.702

.|* |00 .|* 12 0.095 0.143 9.1596 0.689

.|. | .|. |. 13 -0.014 -0.019 9.1816 0.759

.|. | .|. | 14 -0.016 -0.039 9.2132 0.817

.|. | .|. | 15 0.026 0.013 9.2908 0.862

*|. | .|. | 16 -0.092 -0.035 10.308 0.850

*|. | *|. | 17 -0.107 -0.129 11.703 0.818

.|. | *|. | 18 -0.062 -0.101 12.175 0.838

*|. | .|. | 19 -0.100 -0.053 13.422 0.816

.|. | *|. | 20 -0.049 -0.091 13.727 0.844

.|. | .|. | 21 0.062 -0.009 14.223 0.860

.|. | .|. | 22 0.011 -0.042 14.239 0.893

.|. | .|. | 23 0.040 0.016 14.445 0.913

.|. | .|. | 24 0.049 -0.029 14.770 0.927

*|. | *|. | 25 -0.074 -0.081 15.512 0.929

.|. | .|. | 26 -0.047 -0.037 15.813 0.941

.|. | .|. | 27 0.050 0.045 16.158 0.950

.|. | .|. | 28 0.022 0.023 16.223 0.962

.|. | .|. | 29 0.035 0.006 16.401 0.971

.|. | .|. | 30 -0.010 -0.027 16.415 0.979

.|* |00 .|* 31 0.099 0.140 17.863 0.971

.|. | .|. |. 32 0.021 -0.006 17.928 0.979

.|. | .|. | 33 0.049 0.028 18.285 0.982

*|. | *|. | 34 -0.094 -0.089 19.632 0.977

*|. | *|. | 35 -0.136 -0.105 22.506 0.949

.|* | .|. | 36 0.080 0.039 23.528 0.946

And the ACF is different.

Oh, well. I await your conclusions.

 
I have specially prepared a series for you. I count the difference as x(t)/x(t-1) - 1.
 
I use close.
 
alexeymosc:
I have specially prepared a series for you. I count the difference as x(t)/x(t-1) - 1.
I have calculated it. See above.
 
VNG: I cannot figure out the building algorithm.

we take an arbitrary length of the alphabet, in the screenshot it is 24 bits and encode

Red means that price has updated minimum = 1, blue means that price has updated maximum = 0,

And so for each TF. I checked the statement that the trend on the higher TF is "more important", it's partly true, but so far I haven't seen any clear rules

 
faa1947:
Calculated. See above.
I'll do it myself. You're a little hard to understand.
 
alexeymosc:

I'll do it myself. You're a little difficult to understand.
Ready to give explanations.
 

Statistics for series close(t) / close(t-1) - 1:

Statistics from the series close(t) / close(t-1) - 1 rounded to 2 decimal places:

ACF are very similar. But the linear relationships are minimal.

 

Now compare how clearly the correlations between the zero bar and 250 lags are revealed with mutual information. The graph gives a comparison of the quantised series and the random series with the same distribution.

 
alexeymosc:

Now compare how clearly the correlations between the zero bar and 250 lags are revealed with mutual information. The graph gives a comparison of the quantised series and the random series with the same distribution.

What the numbers on the left mean