|
1. Surveys reuslts bias test :
1.1 LiB :
- Residual computed from {$ R = (\hat C_\ell - C_\ell^{th}) / std(\hat C_\ell) \cdot \sqrt{n_{MC}} = (\hat C_\ell - C_\ell^{th}) / std(\langle \hat C_\ell \rangle )$}
LiB | BICEP |
| |
- with naked eye, we already see that LiB seems unbiased, while Bicep is biais mainly for the EE spectrum.
1.2 Biais ?
- Computing the number of point at 1, 2 or 3 sigma, that is to say {$ (|R|< \sigma)/n_{bins} $} with {$ \sigma = 1, 2, 3 $} : expecting the 68-95-99.7 rule.
- 100000 simulations
LiB :
| noise | 1 sigma | 2 sigma | 3 sigma |
R (EE) | 0.10 muK | 1s 0.804 | 2s 0.978 | 3s 1.000 |
R (EE) | 1.00 muK | 1s 0.826 | 2s 0.978 | 3s 1.000 |
R (EE) | 10.00 muK | 1s 0.804 | 2s 0.978 | 3s 1.000 |
R (BB) | 0.10 muK | 1s 0.848 | 2s 0.978 | 3s 1.000 |
R (BB) | 1.00 muK | 1s 0.783 | 2s 0.978 | 3s 1.000 |
R (BB) | 10.00 muK | 1s 0.783 | 2s 1.000 | 3s 1.000 |
- conclusion : No bias found. The 68-95-99.7 rule is pretty well verified.
BICEP :
R (EE) | 0.10 muK | 1s 0.556 | 2s 0.556 | 3s 0.593 |
R (EE) | 1.00 muK | 1s 0.556 | 2s 0.556 | 3s 0.556 |
R (EE) | 10.00 muK | 1s 0.556 | 2s 0.556 | 3s 0.593 |
R (BB) | 0.10 muK | 1s 0.815 | 2s 0.852 | 3s 0.889 |
R (BB) | 1.00 muK | 1s 0.815 | 2s 0.852 | 3s 0.889 |
R (BB) | 10.00 muK | 1s 0.852 | 2s 0.926 | 3s 0.926 |
- conclusion : bias found fo EE. This si due to the theoretical binned spectrum approximation {$ C_b = P_{b \ell} \cdot C_\ell $} and the fact that the EE spectrum is not flat enought.
- conclusion : less bias found fo BB. This si due to the theoretical binned spectrum approximation {$ C_b = P_{b \ell} \cdot C_\ell $}, but the BB spectrum is sufficiently flat.
- mode per mode :
| Bins number | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 |
| Bins val | 8.5 | 22.5 | 36.5 | 50.5 | 64.5 | 78.5 | 92.5 | 106.5 | 120.5 | 134.5 | 148.5 | 162.5 | 176.5 | 190.5 | 204.5 | 218.5 | 232.5 | 246.5 | 260.5 | 274.5 | 288.5 | 302.5 | 316.5 | 330.5 | 344.5 | 358.5 | 372.5 |
R (EE) | 0.10 muK | 9967.8 | 79.9 | -11.0 | -0.8 | -7.1 | -8.2 | -7.6 | -6.5 | -6.4 | -4.3 | 0.1 | 3.3 | 14.6 | 25.3 | 32.5 | 32.6 | 27.1 | 15.5 | 9.3 | 3.0 | -1.2 | -5.0 | -7.9 | -7.9 | -24.6 | 41.8 | -511.8 |
R (EE) | 1.00 muK | 9822.4 | 79.8 | -11.2 | -0.9 | -6.8 | -8.2 | -7.5 | -6.2 | -6.4 | -4.0 | 0.1 | 3.3 | 14.5 | 25.2 | 31.9 | 32.3 | 26.7 | 15.5 | 9.5 | 3.6 | -0.7 | -3.8 | -6.8 | -5.6 | -26.9 | 66.5 | -575.9 |
R (EE) | 10.00 muK | 7827.6 | 76.8 | -12.2 | -0.6 | -5.8 | -7.5 | -7.1 | -5.5 | -5.6 | -3.5 | 0.0 | 2.7 | 12.6 | 20.6 | 25.3 | 26.6 | 20.9 | 13.7 | 8.2 | 3.8 | -0.9 | -2.0 | -4.7 | -2.2 | -24.7 | 71.8 | -407.8 |
R (BB) | 0.10 muK | 37.8 | 3.9 | -2.3 | -0.3 | 0.5 | 0.6 | 0.1 | 2.1 | -0.2 | -1.1 | 1.8 | 0.0 | -1.5 | -2.1 | -0.0 | -2.0 | -0.9 | -3.2 | -2.3 | -3.3 | -3.8 | -6.2 | -8.4 | -2.9 | -33.1 | 91.4 | -585.5 |
R (BB) | 1.00 muK | 31.1 | 4.8 | -2.7 | -0.1 | 0.5 | 0.9 | -0.3 | 2.3 | 0.1 | -1.1 | 1.9 | 0.3 | -1.3 | -1.7 | 0.4 | -2.7 | 0.6 | -3.6 | -1.2 | -1.8 | -2.7 | -5.6 | -4.6 | -2.5 | -23.0 | 41.7 | -211.3 |
R (BB) | 10.00 muK | -46.9 | 28.6 | -10.3 | 3.5 | -0.4 | -0.1 | 0.4 | 1.3 | 0.5 | 0.2 | -2.1 | 1.1 | 0.7 | 0.6 | 0.6 | -1.8 | -0.4 | -0.9 | 0.5 | 0.0 | -0.3 | -3.8 | -1.0 | -2.6 | -4.6 | -1.7 | -17.2 |
- first and last modes have to be neglected when computing a likelihood on 'r'.
2. NaMaster dependence over r :
- namaster behaves badly when r-> 0 :
- Blue : ns=8
- Green : ns=16
- Red : ns=32
- dashed black : BB model
- y-label is wrong, it should be {$ C_\ell $}
r=0.1 | r=0.01 | r=0.001 | r=0.0 |
| | | | |
input EE = TE = 0 | | | | |
- problem solved : mask aposisation incorrect.
3. Pure Tests :
- LiB, _Ns16, Nbins46, Slmax=3, fwhmdeg0.5, r0.001, fsky0.7
- Color solid lines are the error of the spectrum std(Cl)
- Dashed black line is the spectrum model
0.1 muK | | |
1.0 muK | | |
5.0 muK | | |
50.0 muK | | |
4. Estimation and Variance
- Leakage=False means ClEE =0
Survey | EE | BB |
LiB, fsky=0.5 | | |
LiB, fsky=0.6 | | |
BIC, "fsky=0.0159" | | |
5. Vairance ratio :
| fsky=0.5 | fsky=0.6 |
LiB EE | | |
LiB BB | | |
|
|