Hi
I have a panel of administrative data, with several tax years and about 126 million observations. I am investigating the effect of age on retirement income (lump sum retirement income, pension income, and retirement annuity income). I want to use the panel aspect of the data (I have hashed IDs) to do this. One approach could be to interact a continuous age variable with an age categorical variable, which would result graphically in an estimation with a piecewise, discontinuous, linear function (i.e. a function with jumps and straight lines on the graph income vs age). However, in order for me to present the results in a paper more attractively, I would like to estimate a local polynomial for age on income, but I would still like this to follow the panel aspect. The following paper describes a mathematical approach to this: More efficient local polynomial regression with random-effects panel data models , but I would like Stata commands to implement this. I already am familiar with a method in Stata to convert a local polynomial estimation into a semiparametric estimation if need be.
I am requesting that the estimation should use increases in age for individuals, and should not simply pool.
Do you have any ideas about a method to go about this in Stata?
Thanks
I have a panel of administrative data, with several tax years and about 126 million observations. I am investigating the effect of age on retirement income (lump sum retirement income, pension income, and retirement annuity income). I want to use the panel aspect of the data (I have hashed IDs) to do this. One approach could be to interact a continuous age variable with an age categorical variable, which would result graphically in an estimation with a piecewise, discontinuous, linear function (i.e. a function with jumps and straight lines on the graph income vs age). However, in order for me to present the results in a paper more attractively, I would like to estimate a local polynomial for age on income, but I would still like this to follow the panel aspect. The following paper describes a mathematical approach to this: More efficient local polynomial regression with random-effects panel data models , but I would like Stata commands to implement this. I already am familiar with a method in Stata to convert a local polynomial estimation into a semiparametric estimation if need be.
I am requesting that the estimation should use increases in age for individuals, and should not simply pool.
Do you have any ideas about a method to go about this in Stata?
Thanks
