Porportional regression errors
Locked
 
|
I am using SPSS 25.0.02, and a proportional regression. Yesterday, I started encountering two errors when trying to run a model 1. I have one independent variable that is binary (0 or 1) indicator variable. When that variable is in the model, it won't run, telling me that to running a contrast requires a variable with at least two levels. 2. If I exclude the indicator variable, the model wont run, with a reported error of "no model," when I clearly have put predictor variables, and the dependent variable. I checked the archive, and didn't find any threads on this. Thoughts? Gary |
|
|
Are you using the STATS PROPRO REGR extension command or something else? Please post your syntax. On Fri, Apr 12, 2019 at 10:57 AM Gary Rosin <[hidden email]> wrote:
|
Re: Porportional regression errors
Administrator
|
Meanwhile, an error message saying a contrast requires a variable with two
levels suggests to me that for your estimation sample (i.e., the cases used to estimate your model), you have all 0s or all 1s for that variable. Have you checked that? HTH. Jon Peck wrote > Are you using the STATS PROPRO REGR extension command or something else? > Please post your syntax. > > On Fri, Apr 12, 2019 at 10:57 AM Gary Rosin < > grosin@ > > wrote: > >> I am using SPSS 25.0.02, and a proportional regression. >> >> Yesterday, I started encountering two errors when trying to run a model >> >> 1. I have one independent variable that is binary (0 or 1) indicator >> variable. When that variable is in the model, it won't run, telling me >> that >> to running a contrast requires a variable with at least two levels. >> >> 2. If I exclude the indicator variable, the model wont run, with a >> reported error of "no model," when I clearly have put predictor >> variables, >> and the dependent variable. >> >> I checked the archive, and didn't find any threads on this. Thoughts? >> >> Gary >> ===================== To manage your subscription to SPSSX-L, send a >> message to > LISTSERV@.UGA > (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 K Peck > jkpeck@ > > ===================== > To manage your subscription to SPSSX-L, send a message to > LISTSERV@.UGA > (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 [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
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: Porportional regression errors
|
|
I told Bruce yesterday thet the indicator variable was mixed 1s and 0s, but that I had selected only cases with non-missing values of the dependent value.
It then occurred to me that I had not checked to see what csse had been eliminated. Sure enough, somehow I managed to code the selection incorrectly, and had selected *only* cases with missing values. Thus, the "no model" error when running without the indicator variable. As it almost always is, it was operator error! Thanks to Bruce and John. Gary ----- Original Message ----- From: "Bruce Weaver" <[hidden email]> To: "SPSSX-L" <[hidden email]> Sent: Friday, April 12, 2019 4:18:27 PM Subject: Re: Porportional regression errors Meanwhile, an error message saying a contrast requires a variable with two levels suggests to me that for your estimation sample (i.e., the cases used to estimate your model), you have all 0s or all 1s for that variable. Have you checked that? HTH. Jon Peck wrote > Are you using the STATS PROPRO REGR extension command or something else? > Please post your syntax. > > On Fri, Apr 12, 2019 at 10:57 AM Gary Rosin < > grosin@ > > wrote: > >> I am using SPSS 25.0.02, and a proportional regression. >> >> Yesterday, I started encountering two errors when trying to run a model >> >> 1. I have one independent variable that is binary (0 or 1) indicator >> variable. When that variable is in the model, it won't run, telling me >> that >> to running a contrast requires a variable with at least two levels. >> >> 2. If I exclude the indicator variable, the model wont run, with a >> reported error of "no model," when I clearly have put predictor >> variables, >> and the dependent variable. >> >> I checked the archive, and didn't find any threads on this. Thoughts? >> >> Gary >> ===================== To manage your subscription to SPSSX-L, send a >> message to > LISTSERV@.UGA > (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 K Peck > jkpeck@ > > ===================== > To manage your subscription to SPSSX-L, send a message to > LISTSERV@.UGA > (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 [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 ===================== 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 end all this, the STATS PROPOR REGR command omits missing values by default, so there is no need to prescreen the data. On Sat, Apr 13, 2019 at 9:01 AM Gary Rosin <[hidden email]> wrote: I told Bruce yesterday thet the indicator variable was mixed 1s and 0s, but that I had selected only cases with non-missing values of the dependent value. |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |