Документ взят из кэша поисковой машины. Адрес оригинального документа : http://hea-www.harvard.edu/AstroStat/Stat310_0910/jx_20101026.pdf Дата изменения: Tue Jan 26 17:52:40 2010 Дата индексирования: Tue Oct 2 04:17:30 2012 Кодировка: Поисковые слова: п п п п р п р п р п р п

NEW DEM APPROACHES
BY: JIN XU 20/01/2010

Remove the Data Outside of the Disc
X = 1 : 256 Y = [129 - [ 128. 52 - (128.5 - X )2 ] : 128 + [ 128.52 - (128.5 - X )2 ]]

2562 = 65536 pixels changed to 51608.

Figure 1: Original Figure

1


Figure 2: Disk of The Sun

Negative values issue
Name xrt 331 xrt 332 xrt 333 xrt 334 xrt 335 xrt 336 xrt 338 xrt 339 Be med Open.fits Al poly Open.fits C poly Open.fits Al poly Ti poly.fits Be thin Open.fits Al med Open.fits Open Al thick.fits Open Be thick.fits Positive values(Original) 0.6711 0.9700 0.9862 0.9596 0.8974 0.6720 0.4534 0.5489 Positive(Disk) 0.6876 0.9974 0.9982 0.9939 0.9334 0.6928 0.4511 0.5542

The most interesting figure is xrt 338 Open Al thick.fits, the percentage of pixels with positive value is less than 0.5.

New DEM Model
Without consideration of Guassian Random Field
T

Ib time = (
t=1

t Mbt ) time + e

i jb

sd(ei jb ) (Ib time + b ) 2


Where b is computed from the magnitude of the negative values of Ib in the data. (Fit a half normal with mode zero to the negative values in Ib , and use the fitted b .) = T time Ib time = ( t Mbt ) + ei jb ^ Ib time + b Ib time + b t =1 ei jb iid ^

Estimate of
1. Find out all the negative values 2. Get a new vector, V=c(Vn , -Vn ) 3. b = sd(V) ^ Name xrt 331 Be med Open.fits xrt 332 Al poly Open.fits xrt 333 C poly Open.fits xrt 334 Al poly Ti poly.fits xrt 335 Be thin Open.fits xrt 336 Al med Open.fits xrt 338 Open Al thick.fits xrt 339 Open Be thick.fits in the disk of Image b, form a vector Vn

b ^ 0.12 10.2 4.887 2.42 0.274 0.063 0.0602 0.0537

Fit of New Model
Here, the range of Temperature is from 5.5 to 7.5, and the knots of splines are 6, 6.5 and 7.5

3


-1000
5.5

0

1000

2000

beta

3000

4000

5000

6.0

6.5 t

7.0

7.5

Figure 3: Interesting:The trend changed little as I did before, while the scale of Beta changed a lot, about 10 times less.

q

150

200

q

sd

100

q

50

q q q

qq q q q q qqqq qqqqqqqqqqqqqqqqqqqqqqq qq

0
0

10

20 Index

30

40

Figure 4: sd's are quite small

4


Summary for the random fields package
1. Simulation: CondSimu: the function returns conditional simulations of a Gaussian random field GaussRF: These functions simulate stationary spatial and spatio-temporal Gaussian random fields using turning bands/layers, circulant embedding, direct methods, and the random coin method. SimulateRF: Simulation of Random Fields

Figure 5: This is simulated Random Gaussian Field with mean=0, variance=4, nugget=1, scale=10, alpha=1. This covariance model is "stable", that means C (X ) = e x p(- x ), and the total covariance is nugget + variance*cov()/scale. 2. Fit model parameters fitvario: LSQ and Maximum Likelihood Estimation of Random Field Parameters RFparameters: RFparameters sets and returns control parameters for the simulation of random fields

5


3. Kriging:

p 0
0

1

2

3

4

5

6

7

1

2

3 p

4

5

6

7

Figure 6: Before kriging

x 0
0

1

2

3

4

5

6

7

1

2

3 x

4

5

6

7

Figure 7: After kriging

6


4. Regression: Interactive Regression plot 5. other commands that i don't understand.

7