HELP! Panel Estimation

[jmdowns22]

For my class I am finishing a project estimating homicide rates for every state using 3 variables (unemployment, gdp, and gini index)

the model looks like this d(state) = c dlog(stategdp) d(stategini) d(stateunemployment) state(-1) log(stategdp(-1)) stategini(-1) stateunemployment(-1)

however to account for the impact multiplier i was instructed to use this form of the equation:

d(alabama) = c(1) + c(2) *dlog(alabamagdp) + c(3) *d(alabamagini) + c(4) *d(alabamaunemploy) + c(5) *(alabama(-1) + c(6) *log(alabamagdp(-1)) + c(7) *alabamagini(-1) + c(8) *alabamaunemploy(-1))

using alabama for example

however, when trying to use pool to generate a panel of all 50 states when I use the second model accounting for the impact multiplier it generates ' c(5) *(alabama(-1) + c(6) *log(alabamagdp(-1)) + c(7) *alabamagini(-1) + c(8) *alabamaunemploy(-1)) ' all as one variable instead of individually as when I run the equation is a regular regression command.

How can I fix this so that my results are reliable and account for the impact multiplier c(5) ?


thanks

[EViews Gareth]

Not entirely sure I follow your description, but you seem to indicate you are using a pool object. Pool objects do not allow estimation by expression, or nonlinear estimation, which is what you appear to be trying to estimate.

But you also appear to be trying to estimate each state individually, so that begs the question as to why you're using a pool object to begin with.

[jmdowns22]

I want to use the pool function to run the equation for all 50 states to provide a panel of my results for every state, their coefficients, and the significance to report it in a table. However, using the simple model d(state) = c dlog(stategdp) d(stategini) d(stateunemployment) state(-1) log(stategdp(-1)) stategini(-1) stateunemployment(-1) gives unreliable results for the coefficients and respective p-values arent adjusted for the impact multiplier.

Using the model d(state) = c(1) + c(2) *dlog(stategdp) + c(3) *d(stategini) + c(4) *d(stateunemployment) + c(5) *(state(-1) + c(6) *log(stategdp(-1)) + c(7) *stategini(-1) + c(8) *stateunemploy(-1)) adjusts for the impact multiplier as I need it to, however I cannot use the equation in the pool object estimation to provide a panel of results as it recognizes c(5) *(state(-1) + c(6) *log(stategdp(-1)) + c(7) *stategini(-1) + c(8) *stateunemploy(-1)) all as one variable instead of as it would when running it in the command window.

How can adjust the equation so that the coefficients are adjusted for the impact multiplier c(5) as it does when running this model in the command window for panel results?

thanks

[EViews Gareth]

My previous response stands - you cannot use the Pool object to estimate a non-linear expression. You also do not need to use the pool object since you aren't performing pooled estimation.

Use an equation object for each state instead.

[jmdowns22]

There must be a more efficient way to get results for that model for all 50 states than to individually run each one. In addition, I want it in a panel/table to copy and paste into a paper.

[EViews Gareth]

You could write a quick program to loop through the states and estimate each equation. You could then collate the results, or use the EqTabs Add-in to do it for you.