Spearman’s Rank Correlation

by Dr.Hamdi Boukamcha
Spearman’s Rank Correlation [ ru ]
Author: MetaQuotes (2008.02.10 15:46)
Rating: 10
Downloaded: 7983
Spearman's Rank Correlation    SpearmanRankCorr.mq4 (4.8 Kb) View
Spearman’s Rank Correlation is a non-parametric method used in order to make statistical studies of relationship between variables. In this case, the factual degree of parallelism between two numeric sequences will be detected.

The practical calculation of Spearman’s Rank Correlation includes the following stages:

1) pair each indication with its number (rank) and rank them from the highest to the lowest or vice versa;
    2) subtract the two sets of ranks of each pair of values to be compared;
    3) square each difference and add the obtained values;
    4) calculate the rank correlation from the following formula:

Spearman's Rank Correlation   ,
   where Spearman's Rank Correlation   is the sum of squares of rank differences, Spearman's Rank Correlation   is the number of paired observations.
When employing the rank correlation, one conditionally estimates the correlation ratio between indications considering the values being equal to or below 0.3 to be the indications of low correlation ratio, whereas the values between 0.4 and 0.7 are considered to indicate a moderate correlation ratio, values above 0.7 – to indicate a high correlation ratio.
Spearman’s Rank Correlation is a little less powerful than the Parametric Correlation.

It is reasonable to use the rank correlation when there is just a small amount of observations. This method can be used for both numerical data and in the cases, when the registered values are detected by attributes of various intensity. The source of the above description was found here.

Spearman's Rank Correlation

This indicator is one of oscillators. However, being compared to stochastic oscillator, it is smoother. Moreover, it does not delay in pivot points.
The only external parameter that influences the calculation algorithms is  rangeN. It sets the amount of bars, for which we are trying to find regularities. If rangeN = 14, then we take the close price sequence Close[i], Close[i+1], … Close[i+rangeN-1], and construct a rank sequence for them, i.e., we find the location of each close price when the sequence is sorted. In this case, one real chart turns to be compared to another, monotone increasing chart.

The parameter of direction means sorting from the highest to the lowest value (true) or from the lowest to the highest value (false). The value of true shows a more usual picture, whereas false produces an inverted image. Parameter CalculatedBars is introduced in order to limit the amount of bars under calculation, to save CPU resources (though that wasn’t needed). Zero value of this parameter means that calculations will be performed for all history available. Parameter Maxrange = 30 sets the maximum calculation period. This parameter was introduced in order to save resources, as well, so maybe somebody will need it.

3 comments  To post a new comment, please log in or register

I have tried this indicator and it will not even load. I even tried recompiling it, but it will not load on a chart. I’m using MT4 Ver. 4.0, Build 500. Is anyone else having this problem? Is there a fix?
14.07.2013 07:35 memoses

I use this indicator and It has given me VERY NICE RESULTS. Timeframe: 5M, pair: EURUSD.  I trade the first three hours in London and New York markets to ensure volume. I have in mind the levels +0.9 (Overbought) and -0.9 (Oversold), so that when it breaks down after hitting +0.9 SELL, so when it breaks up after hitting -0.9 BUY. NO REPAINTS !!!

Spearman's Rank Correlation


09.01.2013 04:00 dbenjume

Great indicator. It would be great to have a spearman auto-correlation, returning a value between -1 and +1. Being overbought > 0.75 and oversold < -0.75.

Share this article

Leave a comment