| Message/Author |
|
|
| Soo Owen posted on Tuesday, October 02, 2018 - 3:52 pm
|
|
|
Hello,
I'm testing a second-order effect in my LTA model. I used the code below to do it. The remaining part of the code was the same to that of the model with a first-order effect only. The LR test confirmed that the model including the seond-effect fits better to the data. But the output of the second-order model only shows the transition probabilities from C1 to C2 and from C2 to C3. I can't find the transition probabilities from C1 to C3. Where can I find them?
MODEL:
%OVERALL%
C2#1 on C1#1; !time2 on time 1 (first-order)
C2#2 on C1#1;
C2#1 on C1#2;
C2#2 on C1#2;
C3#1 on C2#1; !time3 on time 2 (first-order)
C3#2 on C2#1;
C3#1 on C2#2;
C3#2 on C2#2;
C3#1 on C1#1; !time3 on time 1 (second-order)
C3#2 on C1#1;
C3#1 on C1#2;
C3#2 on C1#2; |
|
|
|
|
| We need to see your full output - send to Support along with your license number. |
|
|
| Soo Owen posted on Tuesday, October 02, 2018 - 6:24 pm
|
|
|
| Thanks for your prompt reply. I actually also tested the invariant transition probabilities, and it turned out that the model is no significant worsening in fit if stationarity was imposed (p = 0.069006). This means that the transition matrices are the exactly same for c1-c2 and c2-c3. Is this case observed in the published articles often? Now, I am a little worried about that my findings are not very interesting or even not publishable due to this stationarity. |
|
|
|
|
| Such stationarity findings are probably quite common. Some LTA (Hidden Markov) algorithms are built on that assumption so it is probably pretty reasonable. |
|
|
| Back to top |
