Quantcast

Graphing multiple data elements with a grouping variable

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Graphing multiple data elements with a grouping variable

cbrawner
Dear SPSS Syntax Experts,

I am trying to graph an independent variable vs. the predictive probability (and 95%CI) from a logistic regression with the lines grouped by gender. I ran the logistic regression via GENLIN because I could save the 95%CI predictive probability data. My plan was to run a scatter plot then fit a quadratic line of best fit to those data and delete the data points. That might not be efficient but it is the method I know. I can do this in Excel/Powerpoint, but was hoping to get 'er done in SPSS. The code shown below does not include a line of best fit. Optimally, I'd like to shade the area between the 95%CIs, but that might be too much to ask. 

Using chart builder I am able to create a scatter plot of the independent variable vs. predictive probability (without 95%CIs) grouped by gender. Code shown below:

**************************************************************************************.
* Plot of independent variable vs. predictive probability grouped by gender.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted Gender MISSING=LISTWISE 
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.exterior(Gender))
END GPL.
**************************************************************************************.

I am also able to create this scatter plot with the 95%CI, but WITHOUT grouping by gender. Code shown below:

**************************************************************************************.
* Plot of independent variable vs. predictive probability (+95%CI); no grouping.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted CIMeanPredictedLower 
    CIMeanPredictedUpper MISSING=LISTWISE REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  DATA: Gender=col(source(s), name("Gender"), unit.category()) 
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  TRANS: pVO2kg_MeanPredicted=eval("pVO2kg - Predicted Value of Mean of Response")
  TRANS: pVO2kg_CIMeanPredictedL=eval("pVO2kg - Lower Bound of CI for Mean of Response")
  TRANS: pVO2kg_CIMeanPredictedU=eval("pVO2kg - Upper Bound of CI for Mean of Response")
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.exterior(pVO2kg_MeanPredicted))
  ELEMENT: point(position(pVO2kg*CIMeanPredictedLower), color.exterior(pVO2kg_CIMeanPredictedL))
  ELEMENT: point(position(pVO2kg*CIMeanPredictedUpper), color.exterior(pVO2kg_CIMeanPredictedU))
END GPL.
**************************************************************************************.

Below is my attempt at combining the code shown above. This resulted in a "GPL error: Translation failed."

**************************************************************************************.
* Plot of independent variable vs. predictive probability (+95%CI) grouped by gender.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted CIMeanPredictedLower 
    CIMeanPredictedUpper Gender MISSING=LISTWISE 
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  TRANS: pVO2kg_MeanPredicted=eval("pVO2kg - Predicted Value of Mean of Response")
  TRANS: pVO2kg_CIMeanPredictedL=eval("pVO2kg - Lower Bound of CI for Mean of Response")
  TRANS: pVO2kg_CIMeanPredictedU=eval("pVO2kg - Upper Bound of CI for Mean of Response")
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.exterior(pVO2kg_MeanPredicted))
  ELEMENT: point(position(pVO2kg*CIMeanPredictedLower), color.exterior(Gender))
  ELEMENT: point(position(pVO2kg*CIMeanPredictedUpper), color.exterior(Gender))
END GPL.
**************************************************************************************.

Thanks for your time.

Best wishes,
Clinton A. Brawner

===================== 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
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Graphing multiple data elements with a grouping variable

Andy W
I think this will technically do what you ask for.

********************************************************************.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted Gender CIMeanPredictedLower CIMeanPredictedUpper  MISSING=LISTWISE
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.interior(Gender))
  ELEMENT: edge(position(region.spread.range(pVO2kg*(CIMeanPredictedLower + CIMeanPredictedUpper))), color.interior(Gender))
END GPL.
********************************************************************.

But I have some different suggestions, especially if the errors overlap, see this blog post, https://andrewpwheeler.wordpress.com/2016/03/08/on-overlapping-error-bars-in-charts/.

************************************************************************************************.
DATA LIST FREE / pVO2kg Gender.
BEGIN DATA
1 1
1 0
2 1
2 0
3 1
3 0
4 1
4 0
5 1
5 0
END DATA.
DATASET NAME Pred.
COMPUTE MeanPredicted = 4 + 1.5*(Gender) + pVO2kg.
COMPUTE CIMeanPredictedLower = MeanPredicted - 2.
COMPUTE CIMeanPredictedUpper = MeanPredicted + 2.
FORMATS pVO2kg Gender (F1.0).
VALUE LABELS Gender 0 'Female' 1 'Male'.
EXECUTE.

*Superimposed directly.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted Gender CIMeanPredictedLower CIMeanPredictedUpper  MISSING=LISTWISE
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  ELEMENT: edge(position(region.spread.range(pVO2kg*(CIMeanPredictedLower + CIMeanPredictedUpper))), color.interior(Gender))
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.interior(Gender))
END GPL.

*Can manually displace some.
COMPUTE NP = pVO2kg - 0.1 + 0.2*Gender.
EXECUTE.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=NP MeanPredicted Gender CIMeanPredictedLower CIMeanPredictedUpper  MISSING=LISTWISE
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("NP"))
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  ELEMENT: edge(position(region.spread.range(pVO2kg*(CIMeanPredictedLower + CIMeanPredictedUpper))), color.interior(Gender))
  ELEMENT: point(position(pVO2kg*MeanPredicted), color.interior(Gender))
END GPL.


*Or treat X as categorical.
GGRAPH
  /GRAPHDATASET NAME="graphdataset" VARIABLES=pVO2kg MeanPredicted Gender CIMeanPredictedLower CIMeanPredictedUpper  MISSING=LISTWISE
    REPORTMISSING=NO
  /GRAPHSPEC SOURCE=INLINE.
BEGIN GPL
  SOURCE: s=userSource(id("graphdataset"))
  DATA: pVO2kg=col(source(s), name("pVO2kg"), unit.category())
  DATA: MeanPredicted=col(source(s), name("MeanPredicted"))
  DATA: Gender=col(source(s), name("Gender"), unit.category())
  DATA: CIMeanPredictedLower=col(source(s), name("CIMeanPredictedLower"))
  DATA: CIMeanPredictedUpper=col(source(s), name("CIMeanPredictedUpper"))
  COORD: rect(dim(1,2), cluster(3,0))
  GUIDE: axis(dim(1), label("pVO2kg"))
  GUIDE: axis(dim(2), label("Predicted Value of Mean of Response"))
  GUIDE: legend(aesthetic(aesthetic.color.exterior), label("Gender"))
  SCALE: cat(aesthetic(aesthetic.color.exterior), include("0", "1"))
  ELEMENT: edge(position(region.spread.range(Gender*(CIMeanPredictedLower + CIMeanPredictedUpper)*pVO2kg)), color.interior(Gender))
  ELEMENT: point(position(Gender*MeanPredicted*pVO2kg), color.interior(Gender))
END GPL.
************************************************************************************************.

Andy W
apwheele@gmail.com
http://andrewpwheeler.wordpress.com/
Loading...