Accepts a variable definition Interval=n where n is the number of urls

over which the regression should be done.  The regression will be calculated
and printed out for each n urls and followed by a regression for the
entire set of data.

tools/footprint/linear-regression.awk

@@ -29,21 +29,58 @@
 # the provisions above, a recipient may use your version of this
 # file under either the MPL or the GPL.
 #
+
+function regress(DATAPOINTS,SX,SY,SXY,SX2)
+{
+   b1 = (DATAPOINTS * SXY - SX * SY) / (DATAPOINTS * SX2 - SX * SX);
+   b0 = (SY - b1 * SX ) / DATAPOINTS;
+   return b1 " * x + " b0; 
+}
+
 BEGIN {
         if (!Skip) Skip = 0;
+        if (Interval) 
+        {
+           Count = 0;
+           IntervalCount = 0;
+        }
       }
 
 NR>Skip {
-     sx += $1;
-     sy += $2;
-     sxy += $1 * $2;
-     sx2 += $1 * $1;
-#     print NR " " sx " " sy " " sxy " " sx2
+        sx += $1;
+        sy += $2;
+        sxy += $1 * $2;
+        sx2 += $1 * $1;
+        #print NR " " sx " " sy " " sxy " " sx2
+
+        if (Interval)
+        {
+           if(Count == Interval-1)
+           {
+              IntervalCount += 1;
+
+              print NR-Count, "-", NR, ":  ", regress(Count,isx,isy,isxy,isx2);
+
+              Count = 0;
+              isx = 0;
+              isy = 0;
+              isxy = 0;
+              isx2 = 0;
+           }
+           else
+           {
+              Count += 1;
+              isx += $1;
+              isy += $2;
+              isxy += $1 * $2;
+              isx2 += $1 * $1;
+           }
+        }
      }
 
 END {
-    DataPoints = NR - Skip;
-    b1 = (DataPoints * sxy - sx * sy) / (DataPoints * sx2 - sx * sx);
-    b0 = (sy - b1 * sx ) / DataPoints;
-    print b1, "* x +", b0; 
+       if(Interval) {
+          print NR-Count, "-", NR, ":  ", regress(Count,isx,isy,isxy,isx2);
+       }
+       print regress(NR-Skip, sx, sy, sxy, sx2); 
     }