Table 1. Results of Determining Size of Spheroidal Scatterers by Fitting Scattering Distribution to Mie Theorya.
| Equator | Polar | Both | Accuracy | ||
|---|---|---|---|---|---|
| Configuration | TE, S11 | 23 | 35 | 13 | 71 |
| TM, S11 | 46 | 17 | 18 | 79 | |
| TM, S22 | 12 | 42 | 19 | 74 | |
| TE, S22 | 67 | 1.0 | 23 | 91 | |
| random, S11 | 37 | 22 | 18 | 77 | |
| randomS22 | 52 | 13 | 24 | 89 | |
| Summary statistics | Cell | 41 | 18 | 18 | 76 |
| Phantom | 39 | 25 | 20 | 84 | |
| S11 | 36 | 24 | 14 | 75 | |
| S22 | 44 | 19 | 21 | 83 | |
| TE | 18 | 39 | 16 | 72 | |
| Random | 45 | 18 | 21 | 83 | |
| TM | 57 | 9.0 | 19 | 85 | |
Data are in percent. The average number of fits to each manifold was assessed for each configuration or summary statistic, where each manifold has a width of 2λ/nS. The final column is a measure of success for the hypothesis; namely, that the size determination is within one wavelength of one of the dimensions of the spheroidal scatterer. Data in boldface correspond to configurations or summary statistics with a normalized root mean square error (NRMSE) less than 0.1; data in lightface correspond to an NRMSE between 0.1 and 0.13. The average null NRMSE is 0.40 and always greater than 0.38. For all configurations and summary statistics, the difference between the NRMSE and the null NRMSE is statistically significant (p<10−3).