03-14-2016 05:51 AM
is there a macro for calculating, e.g. the transition probabilities of matrices that is as elaborate as the R package https://github.com/spedygiorgio/markovchain?
I found papers on Elsevier like doi:10.1016/j.cmpb.2003.12.001, but mostly the download links contained in them are broken.
I wanted to begin with a calculation of transition probabilities for a base-model matrix that looks like this:
OrderID t0 t1 t2
1234 D T D
1235 D B
but extend analysis to higher-order models, consider time-heterogeneity etc.
03-15-2016 06:13 AM
I am also confused by your question. Usually a markov transition model specifies a matrix of probabilities that indicate the transition probabilities between states. You seem to have a character and non-square matrix.
03-15-2016 07:44 AM
The transition matrix I want to estimate is going to be numeric and square. It is going to show the probability of changing from state G to T, for example.
My data matrix has numeric states and does not have to be square.
All I can find are R packages to do so. I am searching for a SAS macro with the same capabilities.
03-15-2016 08:18 AM
Lots of options, depending on what you are modeling and how you choose to model. Do an internet search for
sas markov transition matrix
See the recent paper by Chen (2014): http://www.wuss.org/proceedings14/36_Final_Paper_PDF.pdf
Also the macro by Min, Fang, and Chen (2004): http://www.sciencedirect.com/science/article/pii/S0169260703001391
Contact the authors to obtain the macro.
For alternatives to SAS/IML, see the article http://support.sas.com/kb/24/494.html
03-20-2016 04:07 AM
Thanks for your answer. I already contacted the authors of named paper a while ago, but they don't seem to be active anymore.
All solutions I found in SAS were using numeric data, but I could not find anyone that used character data. Clearly, fitting a logit model etc would not work with my form of dataset.
10-08-2016 04:57 AM
Please provide the current contact details for the corresponding author:
SAS macro program for non-homogeneous Markov process in modeling multi-state disease progression