Re: Compute a weighted correlation



youngjin.michael@xxxxxxxxx wrote:
Hello,

Is there a way to assign weights to data before compute a simple
correlation?
I was googling, but not able to find any answer to this.
Thanks in advance.

Young-Jin

Weighted Mean:
m[x;w] = (sum_i w_i*x_i)/(sum_i w_i)

Weighted Covariance:
c[x,y;w] = (sum_i w_i*(x_i - m[x;w])*(y_i - m[y;w]))/(sum_i w_i)

Weighted Correlation:
r[x,y;w] = c[x,y;w]/sqrt(c[x,x;w]*c[y,y;w])

.



Relevant Pages

  • Re: PCA And Creating Weights For A Composite Measure
    ... of weights for a composite measure related to, e.g., police's ... Collapsing over districts, either before or after forming the ... from between-group correlation and within-group correlation. ... states are the units of analysis you get between-group correlation. ...
    (sci.stat.math)
  • Re: multiple correlations with one click?
    ... Think of two columns of numbers: in col 1 you have body weights and in col 2 ... greater no. of calories per day have higher weights. ... you want to do a correlation between each col and every other col. ... "Otto Moehrbach" wrote: ...
    (microsoft.public.excel.worksheet.functions)
  • Re: hanging cable with discrete loads
    ... Googling found a lot ... Choose smaller weights p1,p2,p3 and tie with knots on a relatively ... relative position and proportion of weights ... Model deflections can ...
    (sci.math)