[ai-geostats] comparisons between classical and robust correlations
I am wondering if you have any thought on this interpretation:
Supposing i have 2 different data sets X and Y. I know that both
data sets are actually a combinations of 2 other data sets, a
background x1 and a contaminating x2 data sets for X and similar
for Y (with y1 and y2).
I encounter 2 cases: a) classical correlation is higher than robust
correlation and b) robust correlation is higher than classical
How i am interpreting the 2 cases? I would like to be able to say
that in case a) the correlation between the contaminating values of
both X and Y increases the overall correlation coefficient, while
when using the robust correlation, the correlation between the
contaminating sets is lost.
Of corse, the reverse is true for case b). The contaminating sets
are not correlated so including them in the calculation of the
classical correlation will decrease this coefficient, while eliminating
these contaminating sets, the robust correlation characterizes the
I will really appreciate any thoughts about the interpretation above.
University of Manchester
School of Geography
Mansfield Cooper Bld. 3.21
Manchester M13 9PL
Tel: +44 (0) 275 8689