Andy Luse posted on Wednesday, October 08, 2014 - 6:05 pm
I am trying to reproduce a dataset using monte carlo simulation that is very close to the original dataset it came from. I first ran the original dataset and saved the estimates. I then used these estimates as the input to my simulation and created and output 1000 different simulated datasets with the same number of observations as the original (using various seeds/etc.)...
TITLE: SCCT LM monte carlo;
MONTECARLO: NAMES ARE intent itse interest career; NOBSERVATIONS = 309; SEED = 9; NREPS = 1000;
POPULATION = estimates_input.dat;
REPSAVE = ALL; SAVE = reps\rep*.dat;
ANALYSIS: TYPE = BASIC;
MODEL POPULATION: intent ON itse interest career;
From here I wrote a program that took these 1000 datasets and averaged each item across all the 1000 datasets to create one dataset with the same number of observations as the original (i.e. averaged row1 column1 for all items in this place across the 1000 datasets, etc.). My thought was that the more datasets I simulate using the original sample estimates, the closer I should get to the parameters in the original sample, but the dataset I am getting is not very close at all. Is there something I am missing?