Nov. 18, 2008
Changed the loop sequence in macro LE_2 in response to Chi-Tsun’s email.
Nov. 24, 2008
Modified the event timing for data with interview gap longer than 1 year. Previous versions record
the events in the last annual interval. Now events other than death occur at midpoint of the gap. If the
gap is odd years, then this is consistent with standard demorgraphic assumptions. But if the gap is even
years, then there is a slight difference. For example, if the observations in year 1 and 3 are A and B.
then the recreated data sequence is A, A and B, effectively delaying the event occurance by 0.5 years.
Mar. 5, 2009 - v.1.03
Modified the sample data set.
Apr. 20, 2009 - v.1.04
Added libname statement to all programs and modified the labnames.
May 8, 2009 - v.1.05
Modified the calculation of life years in the last interval for the simulation programs. Added 0.5 years back to the number of years lived and assigned it to the time in last state.
May 20, 2009 - v.1.06
Modified the error in bootstrap sample selection when only one PSU is present, and the reweighting of sample weight in such case.
SPACE_REVES09 is a special version based on v.04, but incorporated the change made today on May 20, 2009.
June 23, 2009 - v. 2.0
Included a new program that estimates state-duration-dependent transition probabilities based on the SMP-EM model (Cai, Schenker and Lubitz, 2006).
Modified all the SIM1COV and SIM2COV programs such that the new program (SIMxCOV) simulates a single large cohort distributed by the health states and covariates measured at baseline. The old programs simulates a different for each unique combination of covariates, which are inefficient and inflexible. The new program allows users to include any number of covariates.
Modified the timing of events for surveys with gaps of interview larger than 1 year. Previously (see Notes on Nov. 24, 2008) the timing was assumed to be a fixed time point. Based on conversations with Scott Lynch at REVES 21, we decided to randomized the timing. That is, the event time is assumed to occur randomly during the interview gap. For example, if the gap is four years between 1996 and 2000, then an event can occur in 1997, 1998 or 1999; previously it was always assumed to occur in 1998.
Modifed file names and documentations to reflect the above changes.
Renamed all the program files.
Feb. 16, 2010 - v. 2.1
Modified the program to correct a number of minor of mistakes, including
- the program names for MSLT_RAD_S, MSLT_RAD1COV_S and MSLT_RAD2COV_S were not updated in the corresponding M programs
- the data set names in some of the programs were not updated
- the last sections of the MSLT_SIMxCOV_M incorrectly merges the SE output with the point estmates.
- the output data set that holds the simulated trajectories for SMPEM_SIMxCOV_S were not sorted correctly and gives incoore
I also change the list of covariates in MSLT_SIMxCOV_M from SEX RACE EDU to SEX RACE; it is not necessary to do 3 variables for illustration purposes.