Pooling Time Series and Cross Section Data-Some large Sample Properties

Abstract

This thesis deals with the problem of statistical in- ference in combining cross section and time series data. It discusses this problem in different models that have coefficients varying according to two factors.These two factors are individuals and time periods.There is also an assumption about the coefficients.These coefficients might be fixed which leads to .the dummy variable model or seemingly Unrelated regression model or they might be random which leads to the error components models or the Swamy random coefficient model.A statistical inference is made about the last case.The thesis also presents the definition of the K-class estimatr as given by Theil, and an important theorem about its bias to order T-l using

Nagar (1959)'s analysis is introduced.

There is also an extension to Mikhail's result(l97S)

about derivation of the generalized Pooled estimator

where the bias to order. T-l is presented by tlsing ragar's analysis denoted by large T-asymptote.