Hi everyone!
I am new to econometrics and not 100% certain about how to go about analyzing panel data with an interaction term.
I have seen a lot of great comments here and hope that you my very basic questions (apologies) will be answered and that I can get some advice on whether am heading in the right direction.
My dependent variable will be log per capita consumption
My independent variable will be log real GDP per capita and log openness to trade
my control variable will be log inflation rate
sample 7 countries t 20 years
first question is what steps should i take in analyzing panel data. so far i have identified
1. test for Heterosedaskitcty test: (Erlat LM-test) and Serial Autocorrelation: (Baltagi LM-test) among all variables (dependent and independent and control variables) to make sure i don't have high correlation.
2. Run panel unit root test, to determine whether your data is I(0) or I(1). use levin chu, Augmented Dickey Fuller (CADF) statistics or Cross-Sectionally Augmented IPS (CIPS) statistic value allowing cross section dependence.
3. test using hausman to determine whether to run fixed or random effect
4. I then run the dynamic model determined from 3.
Have i covered all the steps? correctly and in the right order?
My second question is that the data collected is in percentage 3% 4.1% 2.27% etc so my data appears as 3, 4.1, 2.27 and so on. do i need to divide all variables by 100 so that i get 0.03, 0.041, 0.0227....or can i just convert it to log without the need to divide by 100
My third question is on interaction terms
the main terms will be the independent variables: real GDP per capita and openness to trade
the interaction terms (real GDP per capita X openness to trade)
do i need to convert the interaction term to log? and do i do that after multiplying them together or i take the log form of GDP and multiply it by the log form if openness to trade
i hope my questions are clear and i hope you can help me out.
thanks alot!
I am new to econometrics and not 100% certain about how to go about analyzing panel data with an interaction term.
I have seen a lot of great comments here and hope that you my very basic questions (apologies) will be answered and that I can get some advice on whether am heading in the right direction.
My dependent variable will be log per capita consumption
My independent variable will be log real GDP per capita and log openness to trade
my control variable will be log inflation rate
sample 7 countries t 20 years
first question is what steps should i take in analyzing panel data. so far i have identified
1. test for Heterosedaskitcty test: (Erlat LM-test) and Serial Autocorrelation: (Baltagi LM-test) among all variables (dependent and independent and control variables) to make sure i don't have high correlation.
2. Run panel unit root test, to determine whether your data is I(0) or I(1). use levin chu, Augmented Dickey Fuller (CADF) statistics or Cross-Sectionally Augmented IPS (CIPS) statistic value allowing cross section dependence.
3. test using hausman to determine whether to run fixed or random effect
4. I then run the dynamic model determined from 3.
Have i covered all the steps? correctly and in the right order?
My second question is that the data collected is in percentage 3% 4.1% 2.27% etc so my data appears as 3, 4.1, 2.27 and so on. do i need to divide all variables by 100 so that i get 0.03, 0.041, 0.0227....or can i just convert it to log without the need to divide by 100
My third question is on interaction terms
the main terms will be the independent variables: real GDP per capita and openness to trade
the interaction terms (real GDP per capita X openness to trade)
do i need to convert the interaction term to log? and do i do that after multiplying them together or i take the log form of GDP and multiply it by the log form if openness to trade
i hope my questions are clear and i hope you can help me out.
thanks alot!
Comment