How to program friction model(kind of cencored regression )
Posted: Thu Jan 13, 2011 10:47 am
Hi,
I am going to program (what is called) the “friction model” in economics and statistics. This model can be used for analysing the government intervention. It looks like tobit but different. I can not deal with this model by using eveiws menu-based command. This model assume that government intervenes in the market only when a certain condition is fulfilled.
For example, INT is the amount of government intervention in the market.
The equation to be estimated is:
(1) INT = a + b*X +ewhere INT indicates (actual) intervention amount and X is control variable.
Let INT* denote the desirable intervention amount – this is the same as the right hand side of equation (1) excluding error terms. Then the acutal value, i.e., INT, have a differnt value depending on whether INT* reach the threshold. That is, government is assumed to intervene only when INT* is larger than the upper positive threshold(p>0) or less than the lower negative threshold(n<0).
That is, there are 3 cases.
(2)
INT = INT* - p + e if INT*>p : intervention
INT= 0 if n<=INT*<=p : no intervention
INT= INT* + n +e if INT*<n : intervention
The parameters(a,b, p, n) in (1) and (2) can be estimated by maximum likelihood. The likelihood function is attached.Note that p and n are expressed in plus and minus delta in attached file.
Is anyone who know how to program this model or has complete code of this model or who can give me a a piece of advice to program this model? Help me.please.
I am going to program (what is called) the “friction model” in economics and statistics. This model can be used for analysing the government intervention. It looks like tobit but different. I can not deal with this model by using eveiws menu-based command. This model assume that government intervenes in the market only when a certain condition is fulfilled.
For example, INT is the amount of government intervention in the market.
The equation to be estimated is:
(1) INT = a + b*X +ewhere INT indicates (actual) intervention amount and X is control variable.
Let INT* denote the desirable intervention amount – this is the same as the right hand side of equation (1) excluding error terms. Then the acutal value, i.e., INT, have a differnt value depending on whether INT* reach the threshold. That is, government is assumed to intervene only when INT* is larger than the upper positive threshold(p>0) or less than the lower negative threshold(n<0).
That is, there are 3 cases.
(2)
INT = INT* - p + e if INT*>p : intervention
INT= 0 if n<=INT*<=p : no intervention
INT= INT* + n +e if INT*<n : intervention
The parameters(a,b, p, n) in (1) and (2) can be estimated by maximum likelihood. The likelihood function is attached.Note that p and n are expressed in plus and minus delta in attached file.
Is anyone who know how to program this model or has complete code of this model or who can give me a a piece of advice to program this model? Help me.please.