Jackknife method for linear regression
f(x) = x + err, x in [-1, 1], err in N(0, 0.1)
Using grid with 0.2 step value by x. Shift object into jackknife method up to 3..10 \sigma. Out of symmetry, use interval from left edge to middle point.
RESULT format.
#---|--- LSE ---|---- JK ----|-- DELTA ---|-- SD_LSE --|--- SD_J ---
0 | 0.99899005 | 0.99898997 | 0.00000817 | 0.10016589 | 0.10016589
----|------------|------------|------------|------------|------------
1 | 0.96913627 | 0.96913992 | 0.00036190 | 0.14047137 | 0.14047137
2 | 0.96976304 | 0.96976625 | 0.00031765 | 0.14061936 | 0.14061936
4 | 0.97079940 | 0.97080181 | 0.00023810 | 0.14063194 | 0.14063194
...etc
Here:
Columns:
LSE parameter value from LSE method for full sample
JK parameter value from Jackknife method
DELTA abs value for difference between LSE and JK
SD_LSEl and
SD_J var. for LSE and JK.
Lines:
First line for original samples.
Next lines for samples with shifted objects. Number is
mean position that has been shifted (counting from left
edge).
For more precision results has been generated 10000 samples with objects count N=100. General result is mean from this.