ROC help!

[mariahdoll]

Hi everyone,

I am a M.Sc. student looking at the habitat suitability of water voles for my research project.

Using the program SPSS (v21) I have computed a series of stepwise logistic regression models to predict their presence and absence in the habitat that I am looking at. They have successfully produced some results but when I try to plot a ROC curve, the following warning message comes up:

Warnings
Either or both of the positive and negative actual state groups have no valid observations. Estimation cannot be performed, and the ROC curve will not be displayed.

I have put the predicted probability values in the 'test variable' box and the variable that was included in the habitat suitability model in the 'state variable' box.

Has anyone had this warning come up before and know how to get round it?

Thank you very much.

Best wishes,

Maria

[Maurice Vergeer]

I believe the problem is in the logistic regression analysis, not the
ROC curve analysis.
I suggest first to try descriptives for the actual and predicted
values before turning to the ROC curve..

furthermore try to use logistic regression avoid the stepwise method
and use the enter method for the final (eg only the final model).

this example should work:
LOGISTIC REGRESSION VARIABLES Y
  /METHOD=enter X1 X2
  /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5)
  /save pred(predY) .
roc predY by Y (1)/print se.

Maurice

On Sun, Aug 4, 2013 at 12:31 PM, mariahdoll
<[hidden email]> wrote:

> Hi everyone,
>
> I am a M.Sc. student looking at the habitat suitability of water voles for
> my research project.
>
> Using the program SPSS (v21) I have computed a series of stepwise logistic
> regression models to predict their presence and absence in the habitat that
> I am looking at. They have successfully produced some results but when I try
> to plot a ROC curve, the following warning message comes up:
>
> Warnings
> Either or both of the positive and negative actual state groups have no
> valid observations. Estimation cannot be performed, and the ROC curve will
> not be displayed.
>
> I have put the predicted probability values in the 'test variable' box and
> the variable that was included in the habitat suitability model in the
> 'state variable' box.
>
> Has anyone had this warning come up before and know how to get round it?
>
> Thank you very much.
>
> Best wishes,
>
> Maria
>
>
>
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/ROC-help-tp5721476.html
> Sent from the SPSSX Discussion mailing list archive at 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

--
________________________________________________
Maurice Vergeer
To contact me, see http://mauricevergeer.nl/node/5
To see my publications, see http://mauricevergeer.nl/node/1
________________________________________________

=====================
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