Qunatile regression warning

[mudskipper]

Hi there :-)

I was asked to do a Qunatile regression as a part of the statistic analysts
in my thesis so for the last few days i'm struggling while trying to use the
QUANTREG via "SPSS- Essentials for R".
No matter which variables i'm addding, I always get this mysterious warning
massage:

In rq.fit.br(x, y, tau = tau, ci = TRUE, ...) : Solution may be nonunique

Sometimes i'm also getting this one:

In rq.fit.br(x, y, tau = tau, ...) : Solution may be nonunique

I do get the coefficients tables and the plot graph but I don't know if I
can use them considering this massages.

Here is an example of a pasted syntax:
SPSSINC QUANTREG DEPENDENT=RQD_RR
  ENTER=total_cholesterol_c QUANTILES = 0.25 0.5 0.75
  /OPTIONS PLOT METHOD = BR STDERR = RANKIID ANOVA=NONE MISSING=LISTWISE
  /SAVE ID=code.

Thank you!



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Jon Peck]

That just means that there might be other equivalent solutions.  You can do a little sensitivity analysis by changing the quantile list slightly, e.g., instead of .5 use .501 and you can try changing the estimation algorithm (under Options) to see if you get about the same results.

On Thu, Feb 28, 2019 at 12:17 PM mudskipper <[hidden email]> wrote:

Hi there :-)

I was asked to do a Qunatile regression as a part of the statistic analysts
in my thesis so for the last few days i'm struggling while trying to use the
QUANTREG via "SPSS- Essentials for R".
No matter which variables i'm addding, I always get this mysterious warning
massage:

In rq.fit.br(x, y, tau = tau, ci = TRUE, ...) : Solution may be nonunique

Sometimes i'm also getting this one:

In rq.fit.br(x, y, tau = tau, ...) : Solution may be nonunique

I do get the coefficients tables and the plot graph but I don't know if I
can use them considering this massages.

Here is an example of a pasted syntax:
SPSSINC QUANTREG DEPENDENT=RQD_RR
  ENTER=total_cholesterol_c QUANTILES = 0.25 0.5 0.75
  /OPTIONS PLOT METHOD = BR STDERR = RANKIID ANOVA=NONE MISSING=LISTWISE
  /SAVE ID=code.

Thank you!



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

--
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[mudskipper]

Oh wow changing it to 0.501 0.751 etc, actually worked! - no more messages
:-)
So just to be sure, is that mean that I can safety use the Qunatile
regression and ignore this message?

Thank you so much Jon!

PS
Perhaps you know about a good tutorial\article that may help me understand
how to interpret the Qunatile regression report?



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Jon Peck]

If the results are pretty much the same with .501, go with either value.  The person who implemented the algorithm told me that .5 was a particularly problematic value for it.

I will send you directly a few articles on quantile regression, since this list does not allow attachments.

On Thu, Feb 28, 2019 at 2:45 PM mudskipper <[hidden email]> wrote:

Oh wow changing it to 0.501 0.751 etc, actually worked! - no more messages
:-)
So just to be sure, is that mean that I can safety use the Qunatile
regression and ignore this message?

Thank you so much Jon!

PS
Perhaps you know about a good tutorial\article that may help me understand
how to interpret the Qunatile regression report?



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

--
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[Bruce Weaver]

Jon, if you post via Nabble (link to this thread given below), you can attach
files.  

http://spssx-discussion.1045642.n5.nabble.com/Qunatile-regression-warning-td5737504.html



Jon Peck wrote
> If the results are pretty much the same with .501, go with either value.
> The person who implemented the algorithm told me that .5 was a
> particularly
> problematic value for it.
>
> I will send you directly a few articles on quantile regression, since this
> list does not allow attachments.





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[mudskipper]

OK Thanks again!



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD