Hello everyone,
I have a question related to specifying max(#) using traj command in Stata.
I am analysing a dependent variable with values ranging from 1 to 9 using a censored normal specification. I wanted to seek your expert guidance on specifying the appropriate “min()” and “max()” values in the model. In particular, I would like to check if you would recommend us to specify:
(a) the theoretical minimum/maximum values of the variable (i.e. min(1) and max(9)); OR
(b) a wider range of values (i.e. min(0) max(10))?
I noticed that the estimates obtained by specifying max(9) vs max(10) are different, and results in different group membership probabilities; however, the estimates obtained by specifying other values for max such as max(10), max(20) or max(100) are similar.
If there are any guidelines you would recommend us to use when selecting the min/max values for the censored normal in the traj command, please advise.
Thank you.
I have a question related to specifying max(#) using traj command in Stata.
I am analysing a dependent variable with values ranging from 1 to 9 using a censored normal specification. I wanted to seek your expert guidance on specifying the appropriate “min()” and “max()” values in the model. In particular, I would like to check if you would recommend us to specify:
(a) the theoretical minimum/maximum values of the variable (i.e. min(1) and max(9)); OR
(b) a wider range of values (i.e. min(0) max(10))?
I noticed that the estimates obtained by specifying max(9) vs max(10) are different, and results in different group membership probabilities; however, the estimates obtained by specifying other values for max such as max(10), max(20) or max(100) are similar.
If there are any guidelines you would recommend us to use when selecting the min/max values for the censored normal in the traj command, please advise.
Thank you.
Comment