[Gretl-users] Visualization of an ordered probit Modell

J. I. j.iseringhausen at web.de
Mon Nov 28 18:36:49 EST 2016


Hey Gretl-Users,

I am kind of a noob in econometrics. For my thesis I had to run a 
regression to determine why people use some spacial product.
My AV is dicrete but ranged (3 steps). So I choosed the ordered porbit 
regression and calculated the MEMs. (could not find a way for AMEs)
All results are fine and there was no problem to interpret them.

And now - my question:
I tried to put all Information from the Gretl Output into a grafic (I 
will attach it)
May you can take a minute an tell me, if it makes sense.

As I mentioned, I am a newbie - so excuse if this is stupid...

You need to know: there are 3 categories but two (the red ones) are the 
"User" and one (green one) "Nonuser". The circles should represent that. 
Sizes (boxes and circles) should represent the total data "n=369".
The 3 Output-categories are the boxes in the circles. MEP-Int and 
MEP-Gel (User) MEP-Nicht (Nonuser). Darker ones are the "Pr=" values for 
the each Output in the first line of the MEMs-Output. As I know, these 
are the predicted probabilitys for each Output, if all regressors are 
held at their means... The lighter ones are the "true" observed data.
Same for the circles. Darker ones for the Output and lighter ones for 
the observed Data.

I my thesis I want to argue, that the prediction just separates the user 
and nonuser, but cant distinguish between the to usergoups too good....
Than I am up to go further and put some more information to the 
regression (information i just got of the users) an do another ordered 
probit with just the two user-Output... an the new model is quite good.

I would apprechiate an answer very much. Please excuse my floppy 
english. Its next to 1 am - and I doing thesis for 2 month day and night 
now!

best gegards J.I.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: Modell_LM_369_ModellC Kopie_2.TIF
Type: image/tiff
Size: 4661260 bytes
Desc: not available
URL: <http://lists.wfu.edu/pipermail/gretl-users/attachments/20161129/ba44c679/attachment-0001.tif>


More information about the Gretl-users mailing list