Report on the Agriculture Survey Data Simulation Program*

In this technical report, we describe a computer program that generates data useful for comparing different estimation methods used to analyze Agriculture Survey data. The generated data is formulated as an intercept model where the intercept represents an interviewer effect. In particular, the simulation study involves adding artificial interviewer error to the Agriculture Survey data. The estimation methods are for the type of data where a primary response variable (y) and a covariate (x) are measured for each sampled unit. The model Yij = f.Jij + bi + (ij is utilized where f.Jij is the true value of the Agriculture Survey response, bi is the systematic errror due to the i th interviewer, and (ij is a random, independent measurement error for all i = 1, ... , I interviewers and j = 1, ... , m units. Thus, the simulated survey response Yij is the sum of the true value, the interviewer variable, and the random error. Additionally, the covariate Xij is defined as Xij = f.Jij + eij, where eij is a random error term used to link the covariate Xij with the true primary response variable f.Jij. The random variables {b;}, {(ij}, and {eij} are mutually independent.