Forum: open-discussion
Monitor Forum | 
Start New Thread
| RE: MVPROBIT [ Reply ] By: Arne Henningsen on 2013-04-21 21:56 | [forum:39529] |
|
Dear Chandra It is basically impossible to predict the time that mvProbit() needs for an estimation. Generally it depends on: - the number of (binary) dependent variables (I guess that the time grows at least in squares with this and you have a quite large number, i.e. 8, which results in 28 correlation coefficients to be estimated) - the number of explanatory variables (you also have a quite large number here) - the correlation between the explanatory variables - the optimisation algorithm - the tolerance limit for detecting "convergence" of the optimisation - the number of observations - the starting values - [...] I suggest to start the empirical analysis with a simple model, i.e with only a very few dependent variables, only a few explanatory variables, and a rather large tolerance for detecting convergence. Then, you can make your mode step-wise more complicated with a longer execution time... Best regards, Arne |
|
| MVPROBIT [ Reply ] By: CHANDRA SEKHAR BAHINIPATI on 2013-04-20 04:01 | [forum:39528] |
|
Hi, I am a new user of R and I am doing MVPROBIT. I have eight dependent variable and 15 independent variable. I have run a MVPROBIT regression. It is very slow. Even after 48 hours, I have not got result. Can you tell me average time required to run MVPROBIT model in R Regards, Chandra |
|
