Problem linking runmplus with M-Plus

[garthr]

I have M-Plus version 4 and M-Plus demo version 7 on my desktop, and Stata version 13. I am teaching a class and using runmplus with M-Plus (students have the version 7 demo installed on their laptops and Stata 13). On my computer, runmplus has no problem locating either version of M-Plus. However, several students are having difficulty getting runmplus to locate M-Plus on their laptops.

I found a response on the listserv indicating that one could assign global macros to point to the location of M-Plus, like this:

global mplusdemo7_path `"C:\Program Files (x86)\Mplus Demo\mpdemo7.exe"'
global mplus_path `"C:\Program Files (x86)\Mplus\Mplus.exe"'

However when I tried this, runmplus could no longer locate either of these versions of M-plus on my computer. I had to close Stata and reopen to restore the defaults, and once again runmplus could locate both versions. I would like to know what I am doing wrong so I can help my students. Any help would be appreciated.

Thanks, Garth

[Sergiy Radyakin]

Garth,

1. check what is the actual location of the program on each computer (tell us). What you found in the internet may differ from your actual setup.
2. try the short paths instead of full paths: global mplusdemo7_path `"C:\Progra~1\MplusD~1\mpdemo7.exe"' some guesswork might be involved in determining "~1" or "~2" etc
3. you are writing that it doesn't work for some students. I figure from this that it does work for the others. Then trace the program on the machines where it works and see what is the setting that it autodetects and compare with ones where it doesn't work. Force the same setting with globals.
4. Check if there is anything in common among the students where it doesnt work (e.g. 64-bit OS, European decimal, non-US version of OS, Windows 8, or some other observable characteristic).
5. The demo version of mplus 7 is available for free from statmodel . Clearly download and installation of mplus should be part of the setup of runmplus.ado.

Best regards, Sergiy Radyakin

[garthr]

Thanks very much for your response. In addition, I think I figured out a way to avoid dealing with the various platform related issues my students may be encountering.

Download or copy the version 7 Mplus demo to a thumb drive then specify the location with a global macro. for example, I installed Mpdemo7.exe directly into my thumbdrive folder E:\MPlus. I got a few error messages related to my computing environment that I ignored and clicked through, and it installed. Then I executed this statement:

global mplus_path `"E:\MPlus\Mpdemo7.exe"' // Note the global is named mplus_path

When I did this I got runmplus to work and locate M-Plus.

Garth

[Sergiy Radyakin]

Garth, note that the new path is a short path (conforming to older DOS requirements). Does it still work if your folder on the thumb drive is e.g. "E:\M Plus\mpdemo7.exe"? (with space, rename both the folder and content of the global) If it stops working - then the problem is long file names. Also note that with USB sticks the letter that it is going to be mounted is hardly predictable. So if your students plug in their own stick first, then the one they get from you, then it might be mounted as F: or something else instead of E:, rendering the setup inoperable again. I'd pursue the HDD installation route to preclude this. Best, Sergiy

[garthr]

Sergiy, thanks, for the feedback. The folder names cannot have a space, you are correct, good catch.

[Phil Schumm]

I'm just curious: With the introduction of commands like gsem, meglm, etc. in Stata 13, the range of analyses that can be done in Mplus but not in Stata is rapidly shrinking. What types of analyses are you teaching that require Mplus instead of Stata?

[Clyde Schechter]

Stata's -sem- and -gsem- do not support categorical latent variables. Though some of these models can be estimated with user-written commands like -cmp- or -gllamm-, I find I still have to fall back on MPlus for most of this kind of work.

[Phil Schumm]

Stata's -sem- and -gsem- do not support categorical latent variables.

Indeed. Discrete latent variables and composite link functions are the two reasons I still occasionally use gllamm. I only asked because since Stata 13 was released, I've had several people ask me about Mplus who didn't realize that the model they wanted to fit could now be fit in Stata. I believe there may be a few other things you can do in Mplus that you can't easily do in Stata (yet), but my guess (perhaps incorrect) is that those tend to be rather advanced things that wouldn't typically appear in most applied classes. Discrete latent variables (e.g., latent-class models, mixture models, etc.), however, are certainly an (important) exception.

[Guest]

For me, another reason to stick with Mplus is Mplus's recent options for Bayes, both in ordinary SEM (with or without categorical variables) and more advanced analyses using Bayes. But for anyone using both Mplus and Stata, I really recommend trying runmplus. It's great.

Guest

[Chen Samulsion]

In my profile.do, I set mplus_path to "D:\Xsoftware\Mplus~7\Mplus.exe"
After running -runmplus-, it gives me error message as: "file __000001.out not found".

. net install runmplus , from(https://s3.amazonaws.com/runmplusbucket) force
. sysuse auto
(1978 Automobile Data)

. runmplus price mpg weight, model(price on mpg weight;)
file __000001.out not found
r(601);