evanbiederstedt

evanbiederstedt

evanbiederstedt

# The -2 log-likelihood
#
# -2lnL \propto m^T C^-1 m + ln det C + N ln (2pi)
#
# First term, m^T C^-1 m is the "model fit term"
# Second term, lndetC is the "complexity penalty"
# Third term, N ln 2pi, a constant
#
# m = tempval
# C = Sij + N_ij

evanbiederstedt

Repeat of https://gist.github.com/evanbiederstedt/a763375f068b05167c26
with vector array of shape (3072,), mean=0, variance=1

tempp = np.random.normal(0.0, 1.0, 3072) # mean = 0, std = 1 = var = 1

# Test 1
vary_x_samples125 = np.logspace(-8, -12, num=40) # C3 parameter, vary from e-08 to e-12
sigma125 = 5e-22 # chose this sigma^2 parameter, hold constant

#FIRST, NO PEAK CODE, i.e. scale covariance matrix by e+21

evanbiederstedt

CODE:
""""
#
# Create an array shaped (3000,), mean = 0.0, variance = 1.0, and compute a_lm values. 
# use np.random.normal(mean, std, size)
# http://docs.scipy.org/doc/numpy/reference/generated/numpy.random.normal.html
#
hoggarray = np.random.normal(0.0, 1.0, 3072) # mean = 0, std = 1 = var = 1

print hoggarray

evanbiederstedt

Same as https://gist.github.com/evanbiederstedt/72b0035ca4b9fce830c3
except vary C3, hold sigma^2 fixed

FIRST, NO PEAK CODE, i.e. scale covariance matrix by e+21
CODE
""""
#
# Hold sigma^2 constant, vary C3
#
vary_x_samples125 = np.logspace(-8, -12, num=40) # C3 parameter, vary from e-08 to e-12

evanbiederstedt

import numpy as np
tempp = (1e6)*tempval # multiply CMB maps by 1e6
print tempp.shape  # array shape

OUTPUT:
(3072,)

print np.median(tempp) # median

OUTPUT:

evanbiederstedt

FIRST, FIND NO PEAK IN LF, NOISE PARAMETERS E-21 TO E-23

CODE:

""""
vary_x_samples125 = np.logspace(-8, -12, num=40) #num = 40
sigma125 = np.logspace(-21, -23, num=40)

Sij = vary_x_samples125[:, None, None] * norm_matrix[1][None, :, :]
newSij = (1e21)*Sij   # multiply S_ij by 1e12

evanbiederstedt

C3_sample1 = 4e-8
sigma2samples1 = np.linspace(1e-22, 6e-23, num=40)
 
 
# param is our parameter, C_3
Sij = C3_sample1 * norm_matrix[1][None, :, :]
newSij = (1e22)*Sij   # multiply S_ij by 1e12
Nij = sigma2samples1[:, None, None] * id_mat[None, :, :]
newNij = (1e22)*Nij
# Format 7/4pi * param * P_3(M) where param is the parameter we vary, C_l

evanbiederstedt

testmattt = Sij[13]
testmattt_inv = np.linalg.inv(Sij[13])

print testmattt

OUTPUT
[[  3.60611595e-10   3.54999009e-10   3.49422963e-10 ...,  -3.54999009e-10
   -3.60611595e-10  -3.54999009e-10]
 [  3.54999009e-10   3.60611595e-10   3.54999009e-10 ...,  -3.49422963e-10
   -3.54999009e-10  -3.60611595e-10]