Module 4 Flashcards
1
Q
2 ways of simming variables
A
- Inverse transform method.
2. Acceptance/Rejection method.
2
Q
Inverse transform method
A
Generate U ~ unif(0,1) then feed this into the inverse of the CDF of summing variable.
Ie put in the y values of the CDF and extra the observations x
3
Q
When is acceptance/reject method good
A
When we cant invert the cdf
4
Q
Acc/Rej method, f(x) represents
A
what we want to simulate
5
Q
Acc/Rej method, g(x) represents
A
what we know, and will use to assist.
6
Q
Condition for c in Acc/Rej method
A
f(y) <= g(c)*c
7
Q
Condition for acceptance Acc/Rej
A
U <= f(x) / c*g(x)
8
Q
2 Methods of variance reduction
A
- Antithetic Variables (average 2 samples with -ve correlation)
- Control Variates
9
Q
Why does -ve correlation reduce variance…
A
Look at Var[ (Y1+Y2)/2 ],
it simplifies to Var(Yboth) = Var(Y)(1+p)/K,
so if p is -ve, var is reduced.
