warning: variance matrix is nonsymmetric or highly singular

[Rabab Al hasni]

Dear all,

I have performed multinomial regression that is comprised of a dependent variable (4 categories; coding starts from 0 to 3) and 32 independent variables of company characteristics and owners. However, when it comes to estimating the marginal effect average for each category of the dependent variables I got for the third category STATA software warning which says: 'variance matrix is nonsymmetric or highly singular' I searched to understand and solve the problem but could not find a clear solution. An example of the problem is attached.

Please, could you advise me what to do in this situation? How to solve it please?

My research objective is to identify the determinants of the factors that would affect approaching bank financing. Therefore and in case if the problem of the variance matrix not solve, can I use the estimated coefficient s to interpret the analysis instead of the average marginal effect estimation?




Best regards,
Rabab

[Carlo Lazzaro]

Rabab:
you might be interested in the following thread: https://www.statalist.org/forums/for...ighly-singular.
As this one is not your first post on this forum, you should be already aware of that attachments are discouraged (please, read and act on the FAQ), let alone spreadsheets, that no one on this list would ever download, due to the risk of active contents.

[Rabab Al hasni]

Rabab:
you might be interested in the following thread: https://www.statalist.org/forums/for...ighly-singular.
As this one is not your first post on this forum, you should be already aware of that attachments are discouraged (please, read and act on the FAQ), let alone spreadsheets, that no one on this list would ever download, due to the risk of active contents.

Thank you Carlo for bringing my attention and awareness regarding the attachment. is it possible to delete it from my post?

Appreciate your concern

Best regards,
Rabab

[Rabab Al hasni]

Rabab:
you might be interested in the following thread: https://www.statalist.org/forums/for...ighly-singular.
As this one is not your first post on this forum, you should be already aware of that attachments are discouraged (please, read and act on the FAQ), let alone spreadsheets, that no one on this list would ever download, due to the risk of active contents.

Many thanks Carlo for your prompting reply and posting the above link. I have tried to apply your advice regarding the MandatedAssurance in Stata 15 but did not recognize it. Do I need to install SSC specific dataex that MandatedAssurance in my Stata software? In fact, I typed this "MandatedAssurance (unlabeled)" as a command in the Stata software but I got the error.

Regarding the attachment, I apologize for this mistake. I have tried to solve but I think I it does not allow for editing now. I would appreciate if you could advise on how to remove or delete the attachment.


Appreciate your kind efforts and concern

Best regards
Rabab

[Carlo Lazzaro]

Rabab:
-MandatedAssurance- is not a Stata command, but the name of one of the variables included in the dataset of the original poster of the thread I quoted.
In your case, you should find out the variable(s) that may cause the variance matrix to be nonsymmetric or highly singular.

[Jean-Claude Arbaut]

Carlo Lazzaro

I generally agree with your comment on attachments, however regarding the risk of active contents, see this:
https://superuser.com/questions/6424...e-safe-to-open

I would add that not only will Excel refuse to open an xlsx file if there is a macro (which shouldn't even be possible, but one could try to hack the file),
but Excel will only open file from the web in protected mode, if configured correctly (the default).
And most other programs, including Stata, won't ever attempt to execute macros.
So, basically, xlsx is safe. Of course, it's not true of xls, xlsm, xlsb, xla or xlam files if you open them with Excel.

[Rabab Al hasni]

Rabab:
-MandatedAssurance- is not a Stata command, but the name of one of the variables included in the dataset of the original poster of the thread I quoted.
In your case, you should find out the variable(s) that may cause the variance matrix to be nonsymmetric or highly singular.

Thank you so much Carlo for the clarification.

Appreciate your kind efforts


Best regards,
Rabab

[Rabab Al hasni]

Dear Jean-Claude and Carlo,

I am really sorry I am not aware of the files attachments issues. I apologize again if the attachment may cause a problem. I would like to delete it or remove from my post but I could not find an option in the page to do so. If Stata can do this I would greatly appreciate it.

So I can understand that I should not post excel or file attachment in the forum. Many thanks for bringing to me this attention for future considerations.

Best regards,
Rabab

[Carlo Lazzaro]

Rabab:
see the FAQ about CODE delimiters and -dataex- to share with interested listers what you're doing. Thanks.