Question about "measure column" in variable view
Locked
 
|
Hi,
I have a question about the “measure column” in variable view in SPSS. I know that measurement levels are important and that it is important to be aware of the measurement level of a variable when doing statistical analysis in SPSS. It is a key thing. But is it really that important in SPSS to decide whether a variable should be set to nominal, ordinal or scale in the measure column in the variable view? When will that choice in the measure column influence the statistical analysis that I order? My SPSS instructor at my university always taught me that I needed to decide the measurement level of a variable myself and not to trust SPSS in that matter. From what I can see, whether a variable is set to nominal, ordinal or scale in the measure column does not influence results in a regression analysis etc. So I trust myself and not SPSS. But when is it important to set the proper measurement level to either nominal, ordinal or scale before doing statistical analysis in SPSS? I guess the measure column is there for reason? Thanks for helping me to understand this! Best regards, Mimir -- 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 |
Re: Question about "measure column" in variable view
|
I always change the column attribute order and put ML first. See If you use data from archives, you will find that automatic archiving software often changes all the levels. Best to set your own. John F Hall MA (Cantab) Dip Ed (Dunelm) IBM-SPSS Academic Author 9900074 Email: [hidden email] Website: https://surveyresearch.weebly.com/ Course: https://surveyresearch.weebly.com/1-survey-analysis-workshop-spss.html -----Original Message----- Hi, I have a question about the “measure column” in variable view in SPSS. I know that measurement levels are important and that it is important to be aware of the measurement level of a variable when doing statistical analysis in SPSS. It is a key thing. But is it really that important in SPSS to decide whether a variable should be set to nominal, ordinal or scale in the measure column in the variable view? When will that choice in the measure column influence the statistical analysis that I order? My SPSS instructor at my university always taught me that I needed to decide the measurement level of a variable myself and not to trust SPSS in that matter. From what I can see, whether a variable is set to nominal, ordinal or scale in the measure column does not influence results in a regression analysis etc. So I trust myself and not SPSS. But when is it important to set the proper measurement level to either nominal, ordinal or scale before doing statistical analysis in SPSS? I guess the measure column is there for reason? Thanks for helping me to understand this! Best regards, Mimir -- 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 |
|
|
In reply to this post by Odin
Most procedures in SPSS do not use the measurement level in determining how to treat the data. It is mostly an advisory device via the variable icons as a reminder to the user. However, some procedures do use the ML actively. This includes TREES, CTABLES, and the Chart Builder. And the R-based extension commands, of which there are over 50, do use the ML to determine whether to treat variables as factors or not. There are heuristics built in to Data > Define Variable Properties and some other places. For example, if the variable has a currency format, it is probably scale, but it is always better for the analyst to set the ML with knowledge of what the data mean. In some cases, you might want to use a variable both as categorical and scale. On Wed, Apr 7, 2021 at 6:41 AM Mimir <[hidden email]> wrote: Hi, |
|
|
In reply to this post by John F Hall
Hi, and thanks for the answer.
I must admit that I have never changed the specified measurement levels for the variables in my SPSS files in the “measure column” (that SPSS may have automatically decided). I know the measurement level of the variables in my data and I use the variables accordingly in statistical analysis. For example, a simple correlation analysis with two variables will yield the same pearsons r correlation coefficient it I have set the variables to “nominal” or to “scale” in the “measure column”. Therefore I have not given much attention to what the “measure column” says. But maybe that is wrong? I am getting a bit anxious now that I may have produced bad results by not giving much attention to what the “variable column” says…Any insights into when it is absolutely imperative to have the correct specifications in the variable column? Best, Mimir -- 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 |
|
|
In reply to this post by Jon Peck
Thank you so much for your answer! Best, Mimir
-- 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 |
Re: Question about "measure column" in variable view
|
In reply to this post by Jon Peck
In my experience, heuristics are just that heuristic.
I am a firm believer in thorough quality assurance on the data definition in the data view. Part of that QA is that the user understands what the data are. I rarely change the measurement level GUESSED (aka heuristically derived) measurement level. The most frequent occasion is when a numeric variable is guessed as scale level when it should be nominal level. Of course, if it is nominal level it should have value labels. The presence of a nominal level variable without value labels usually warns us the data definition is not yet complete and running procedures is very questionable. Of course, 2-valued variables (aka dichotomies, flag variables) can often be thought of as interval because there is only 1 interval and therefore it is equal to itself. ----- Art Kendall Social Research Consultants -- 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
Art Kendall
Social Research Consultants |
Re: Question about "measure column" in variable view
|
|
In reply to this post by Odin
Hi, Mimir, a tiny example is how CTABLES works with variables. If it's a scale it "automatically" calculates the mean while non scale variables are interpreted as categorical and frequencies are shown. That might be a reason. Good luck, Mario Giesel Munich, Germany
Am Mittwoch, 7. April 2021, 14:41:32 MESZ hat Mimir <[hidden email]> Folgendes geschrieben:
Hi, I have a question about the “measure column” in variable view in SPSS. I know that measurement levels are important and that it is important to be aware of the measurement level of a variable when doing statistical analysis in SPSS. It is a key thing. But is it really that important in SPSS to decide whether a variable should be set to nominal, ordinal or scale in the measure column in the variable view? When will that choice in the measure column influence the statistical analysis that I order? My SPSS instructor at my university always taught me that I needed to decide the measurement level of a variable myself and not to trust SPSS in that matter. From what I can see, whether a variable is set to nominal, ordinal or scale in the measure column does not influence results in a regression analysis etc. So I trust myself and not SPSS. But when is it important to set the proper measurement level to either nominal, ordinal or scale before doing statistical analysis in SPSS? I guess the measure column is there for reason? Thanks for helping me to understand this! Best regards, Mimir -- ===================== 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 |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |