The weighting efficiency provides the amount of distortion that was needed to arrive at the weighted figures. I.E how much the data is manipulated by the weighting.
The result is stored in the report of weighting located into the survey.dat directory
A low efficiency indicates a larger bias introduced by the weights.
You'll find it into the weighting's report and below the formula to calculate it :
For eg.
| w i | w i² | |
| i1 | 0.820508273 | 0.67323383 |
| i2 | 1.027660121 | 1.05608532 |
| i3 | 0.798126889 | 0.63700653 |
| i4 | 1.813092082 | 3.2873029 |
| i5 | 0.524670726 | 0.27527937 |
| i6 | 1.682138128 | 2.82958868 |
| i7 | 0.066485204 | 0.00442028 |
| i8 | 1.524746555 | 2.32485206 |
| i9 | 0.571220139 | 0.32629245 |
| i10 | 1.878748796 | 3.52969704 |
| i11 | 0.070238083 | 0.00493339 |
| i12 | 1.840453697 | 3.38726981 |
| i13 | 0.683791839 | 0.46757128 |
| i14 | 1.00838961 | 1.01684961 |
| i15 | 0.833383964 | 0.69452883 |
| i16 | 1.330310173 | 1.76972516 |
| i17 | 0.602917671 | 0.36350972 |
| i18 | 1.257201562 | 1.58055577 |
| i19 | 0.632265608 | 0.3997598 |
| i20 | 1.451447184 | 2.10669893 |
| i21 | 0.622107103 | 0.38701725 |
| i22 | 1.252551275 | 1.5688847 |
| i23 | 0.966905785 | 0.9349068 |
| i24 | 1.561603913 | 2.43860678 |
| Sum | 24.82096438 | 32.0645763 |
| sum(wi) | sum(wi²) |
| n | 24 |
| sum(wi) | 24.82096438 |
| sum(wi)² | 616.0802728 |
| sum(wi²) | 32.06457627 |
| WE | = 616.080273 / (24 * 32.0645763) |
| WE |
80.06 % The weighting reduces the reliability of the sample of 19.1% (~100 - 80.06 ~) In other words , it's like we removed 5 persons from our sample (24 X 80.06 % =19) |
Our sample is matching to 80.06% of the population and the effective base is equal to 19
When the weighting efficiency is below to 80% , means that you have some mismatch between the sample and the population.
And when is below to 70% , we recommend you to check the weight design.
Comments