Heterogeneous Agent Model

[obaid610]

Dear all,

I am trying estimate a Heterogeneous Agents model that forecasts asset price changes. Two types of investors switch between strategies, either chartist or fundamentalist. The relative weights of the agents are determined dynamically by weight W, which is based on relative past profitability of the strategies. y is a switching parameter,which determines the intensity of switching. Af and Ac are performance measures of the fundamentalist and chartist respectively. I need to estimate the parameters c1, c2, the switching parameter y, and the optimal length of K.

Equation 1 shows how changes in asset price are determined. It says that changes in next period are determined by weights of fundamentalists expecting the price to return to fundamental price F, and chartist expecting the trend Pt-(P-1) to extrapolate.
Equation 2 is how weights are determined inside the model. This needs to be generated for a best fit.
3 and 4 are profitability measures of both strategies.



1. Delta P(t+1)= W(t)*{c1[P(t)-f(t)]}+(1-W)*{c2*[P(t)-p(t-1)]}+ error term
2. W(t)={1+exp[-y(Af(t)-Ac(t))/(Af(t)_Ac(t))]}^-1
3. Af(t)=-SUM{c1*{[p(t-k)-f(t-k)]-delta P(t-k+1)}^2}
4. Af(c)=-SUM{c2*{[(p(t-k)-p(t-k-1))]-delta P(t-k+1)}^2}

W is a weight parameter with values between 0-1.

I would be deeply grateful if someone can help me estimate this model. I want to make two estimations, one where switching is allowed, and one where it is not and the two type agents stay balanced throughout with weights 0.5 both.
If additional info is required, please let me know!

[trubador]

viewtopic.php?f=4&t=275

[obaid610]

Dear trubador

Thank you for your reply. This appears to be a simulation whereas I am looking for an estimating the entire system at once.

A little research on the forums has indicated that LogL object can be used to used to estimate this system. Can you advise me on how this can be estimated using logL object.

I will greatly appreciate any assistance with this

Cheers