Selecting the best unbalanced repeated measures model
This paper examined several indices of goodness of fit that available from some statistics software. The authors first introduced repeated data that collected from different time point. As data are not collect at once, it is possible that some data point may be missing. Three types of missing data were briefly explained: missing complete at random (MCAR), missing at random (MAR), and missing not at random (MNAR). MCAR means that missing data are independent both of observed and unobserved outcomes of the variable being analyzed; MAR is missing data occurring in follow-ups made after the treatment phase; and data is nether MCSR nor MAR is MNAR. MAR is a realistic mechanism for most practical applications. A number of conditions were conducted to examinee whether specific fit index show consistently good in selecting true model across different independent variables. Results indicated that no specific fit index have the best fit consistently for every condition. Moreover, it was found that the more the time points, the more the sample size, the more precision would be.