topic Re: Plot RR (95% CI) paneled grid plots in Graphics Programming

Plot RR (95% CI) paneled grid plots

Cruise — Fri, 19 Jun 2020 02:06:34 GMT

Hi Folks:

I'm trying to visualize the RR (95% CI) stratified by three time periods (early, middle and late) comparing between two subsets of data including Detroit vs excluding Detroit.

I have tables. But they're very busy. Is it possible to create a plot as shown below using the data I posted in the code snippet? Help creating this paneled forest plots is greatly appreciated.

data ci;
input Phases $ Detroit $ Variables $ rr	rr_l rr_u p_value;
cards;
early	yes	Var1	1.0524	1.00966	1.09695	0.0158
early	yes	Var2	0.84551	0.79209	0.90254	0.0001
early	yes	Var3	1.0543	1.03979	1.06902	0.0001
early	yes	Var4	0.86697	0.81497	0.9223	0.0001
early	yes	Var5	1.00921	0.98268	1.03646	0.5
early	yes	Var6	0.70181	0.58942	0.83561	0.0001
early	yes	Var7	1.30173	1.03098	1.64358	0.0267
middle	yes	Var1	1.04559	1.01123	1.08111	0.0089
middle	yes	Var2	0.85085	0.80236	0.90227	0.0001
middle	yes	Var3	1.047	1.02776	1.0666	0.0001
middle	yes	Var4	0.92206	0.8788	0.96745	0.0009
middle	yes	Var5	0.99772	0.97841	1.01741	0.8189
middle	yes	Var6	0.97695	0.84688	1.12701	0.7491
middle	yes	Var7	0.98328	0.85264	1.13394	0.8167
late	yes	Var1	1.04109	1.0112	1.07187	0.0067
late	yes	Var2	0.97512	0.93336	1.01875	0.2593
late	yes	Var3	1.00412	0.9892	1.01926	0.5906
late	yes	Var4	0.99671	0.95982	1.03503	0.8642
late	yes	Var5	0.98736	0.97207	1.00289	0.1102
late	yes	Var6	1.06743	0.93955	1.2127	0.3162
late	yes	Var7	0.81891	0.73103	0.91736	0.0006
early	no	Var1	0.99983	0.97043	1.03011	0.991
early	no	Var2	0.96015	0.90744	1.01592	0.158
early	no	Var3	1.03307	1.01981	1.0465	0.0001
early	no	Var4	0.90236	0.86039	0.94637	0.0001
early	no	Var5	0.9896	0.97165	1.00789	0.2633
early	no	Var6	1.09426	0.90321	1.32571	0.3575
early	no	Var7	1.05676	0.89367	1.24961	0.5186
middle	no	Var1	1.02193	0.99727	1.04721	0.0818
middle	no	Var2	0.93691	0.89514	0.98062	0.0051
middle	no	Var3	1.01948	1.00417	1.03503	0.0125
middle	no	Var4	0.94738	0.91323	0.9828	0.0039
middle	no	Var5	0.99374	0.98005	1.00763	0.3751
middle	no	Var6	1.21477	1.06634	1.38385	0.0034
middle	no	Var7	0.89446	0.80586	0.9928	0.0361
late	no	Var1	1.03753	1.00869	1.06718	0.0104
late	no	Var2	0.99744	0.95469	1.04211	0.9089
late	no	Var3	0.99716	0.98166	1.0129	0.7216
late	no	Var4	1.00295	0.96614	1.04116	0.8775
late	no	Var5	0.98597	0.97142	1.00075	0.0627
late	no	Var6	1.13068	0.9913	1.28964	0.0673
late	no	Var7	0.79133	0.7083	0.8841	0.0001
;

Re: Plot RR (95% CI) paneled by vertical and horizontal dimension / forest plots

Cruise — Thu, 18 Jun 2020 21:39:59 GMT

CI95% for variable 6 and 7 are lot wider than the rest. So these two may cause problem requiring to adjust the axis scales.

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Fri, 19 Jun 2020 12:38:12 GMT

data ci;
infile cards expandtabs;
input Phases $ Detroit $ Variables $ rr	rr_l rr_u p_value;
cards;
early	yes	Var1	1.0524	1.00966	1.09695	0.0158
early	yes	Var2	0.84551	0.79209	0.90254	0.0001
early	yes	Var3	1.0543	1.03979	1.06902	0.0001
early	yes	Var4	0.86697	0.81497	0.9223	0.0001
early	yes	Var5	1.00921	0.98268	1.03646	0.5
early	yes	Var6	0.70181	0.58942	0.83561	0.0001
early	yes	Var7	1.30173	1.03098	1.64358	0.0267
middle	yes	Var1	1.04559	1.01123	1.08111	0.0089
middle	yes	Var2	0.85085	0.80236	0.90227	0.0001
middle	yes	Var3	1.047	1.02776	1.0666	0.0001
middle	yes	Var4	0.92206	0.8788	0.96745	0.0009
middle	yes	Var5	0.99772	0.97841	1.01741	0.8189
middle	yes	Var6	0.97695	0.84688	1.12701	0.7491
middle	yes	Var7	0.98328	0.85264	1.13394	0.8167
late	yes	Var1	1.04109	1.0112	1.07187	0.0067
late	yes	Var2	0.97512	0.93336	1.01875	0.2593
late	yes	Var3	1.00412	0.9892	1.01926	0.5906
late	yes	Var4	0.99671	0.95982	1.03503	0.8642
late	yes	Var5	0.98736	0.97207	1.00289	0.1102
late	yes	Var6	1.06743	0.93955	1.2127	0.3162
late	yes	Var7	0.81891	0.73103	0.91736	0.0006
early	no	Var1	0.99983	0.97043	1.03011	0.991
early	no	Var2	0.96015	0.90744	1.01592	0.158
early	no	Var3	1.03307	1.01981	1.0465	0.0001
early	no	Var4	0.90236	0.86039	0.94637	0.0001
early	no	Var5	0.9896	0.97165	1.00789	0.2633
early	no	Var6	1.09426	0.90321	1.32571	0.3575
early	no	Var7	1.05676	0.89367	1.24961	0.5186
middle	no	Var1	1.02193	0.99727	1.04721	0.0818
middle	no	Var2	0.93691	0.89514	0.98062	0.0051
middle	no	Var3	1.01948	1.00417	1.03503	0.0125
middle	no	Var4	0.94738	0.91323	0.9828	0.0039
middle	no	Var5	0.99374	0.98005	1.00763	0.3751
middle	no	Var6	1.21477	1.06634	1.38385	0.0034
middle	no	Var7	0.89446	0.80586	0.9928	0.0361
late	no	Var1	1.03753	1.00869	1.06718	0.0104
late	no	Var2	0.99744	0.95469	1.04211	0.9089
late	no	Var3	0.99716	0.98166	1.0129	0.7216
late	no	Var4	1.00295	0.96614	1.04116	0.8775
late	no	Var5	0.98597	0.97142	1.00075	0.0627
late	no	Var6	1.13068	0.9913	1.28964	0.0673
late	no	Var7	0.79133	0.7083	0.8841	0.0001
;
proc format;
value $fmt
'yes'='Inclde Detroit'
'no' ='exclude Detroit';
run;
proc sgpanel data=ci noautolegend;
styleattrs datacontrastcolors=(yellow orange red);
panelby  Detroit Variables/ layout=lattice onepanel novarname proportional  
rowheaderpos=left noheaderborder colheaderpos=top  spacing=4   ;
scatter x= Phases y=rr/ markerattrs=(symbol=diamondfilled) group=Phases;
highlow x=Phases low=rr_l high=rr_u/group=Phases;
refline 1/axis=y lineattrs=(pattern=dash);
colaxis  label=' '  ;
rowaxis  label=' ' values=(1.5 1 0.5) ;
format Detroit $fmt.;
run;

Re: Plot RR (95% CI) paneled grid plots

Cruise — Mon, 22 Jun 2020 10:42:46 GMT

@Ksharp

Hi Ksharp. Sorry for a delay.

Thank you very much for the solution.

The plot below is the result from R. This plot mixes all in one panel and value of y-axis is ascending. I want it descending like you did. Your SAS approach splits All-data vs excluding Detroit making comparison across row easier.

Ksharp, can you help modify your code to show 95% bars with small range and closer to 1 look more prominent. For example, the plot below shows the upper and lower confidence limit even for the Var 3 despite its tightness. Can this be done by putting Var 6 and Var7 on different scale? and Var1-Var5 on different scale?

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Mon, 22 Jun 2020 11:15:45 GMT

OK. Here is.

data ci;
infile cards expandtabs;
input Phases $ Detroit $ Variables $ rr	rr_l rr_u p_value;
cards;
early	yes	Var1	1.0524	1.00966	1.09695	0.0158
early	yes	Var2	0.84551	0.79209	0.90254	0.0001
early	yes	Var3	1.0543	1.03979	1.06902	0.0001
early	yes	Var4	0.86697	0.81497	0.9223	0.0001
early	yes	Var5	1.00921	0.98268	1.03646	0.5
early	yes	Var6	0.70181	0.58942	0.83561	0.0001
early	yes	Var7	1.30173	1.03098	1.64358	0.0267
middle	yes	Var1	1.04559	1.01123	1.08111	0.0089
middle	yes	Var2	0.85085	0.80236	0.90227	0.0001
middle	yes	Var3	1.047	1.02776	1.0666	0.0001
middle	yes	Var4	0.92206	0.8788	0.96745	0.0009
middle	yes	Var5	0.99772	0.97841	1.01741	0.8189
middle	yes	Var6	0.97695	0.84688	1.12701	0.7491
middle	yes	Var7	0.98328	0.85264	1.13394	0.8167
late	yes	Var1	1.04109	1.0112	1.07187	0.0067
late	yes	Var2	0.97512	0.93336	1.01875	0.2593
late	yes	Var3	1.00412	0.9892	1.01926	0.5906
late	yes	Var4	0.99671	0.95982	1.03503	0.8642
late	yes	Var5	0.98736	0.97207	1.00289	0.1102
late	yes	Var6	1.06743	0.93955	1.2127	0.3162
late	yes	Var7	0.81891	0.73103	0.91736	0.0006
early	no	Var1	0.99983	0.97043	1.03011	0.991
early	no	Var2	0.96015	0.90744	1.01592	0.158
early	no	Var3	1.03307	1.01981	1.0465	0.0001
early	no	Var4	0.90236	0.86039	0.94637	0.0001
early	no	Var5	0.9896	0.97165	1.00789	0.2633
early	no	Var6	1.09426	0.90321	1.32571	0.3575
early	no	Var7	1.05676	0.89367	1.24961	0.5186
middle	no	Var1	1.02193	0.99727	1.04721	0.0818
middle	no	Var2	0.93691	0.89514	0.98062	0.0051
middle	no	Var3	1.01948	1.00417	1.03503	0.0125
middle	no	Var4	0.94738	0.91323	0.9828	0.0039
middle	no	Var5	0.99374	0.98005	1.00763	0.3751
middle	no	Var6	1.21477	1.06634	1.38385	0.0034
middle	no	Var7	0.89446	0.80586	0.9928	0.0361
late	no	Var1	1.03753	1.00869	1.06718	0.0104
late	no	Var2	0.99744	0.95469	1.04211	0.9089
late	no	Var3	0.99716	0.98166	1.0129	0.7216
late	no	Var4	1.00295	0.96614	1.04116	0.8775
late	no	Var5	0.98597	0.97142	1.00075	0.0627
late	no	Var6	1.13068	0.9913	1.28964	0.0673
late	no	Var7	0.79133	0.7083	0.8841	0.0001
;
proc format;
value $fmt
'yes'='Inclde Detroit'
'no' ='exclude Detroit';
run;
proc sgpanel data=ci noautolegend;
styleattrs datacontrastcolors=(yellow orange red);
panelby  Detroit Variables/ layout=lattice onepanel novarname proportional  
rowheaderpos=left noheaderborder colheaderpos=top  spacing=4 uniscale=column  ;
scatter x= Phases y=rr/ markerattrs=(symbol=squarefilled) group=Phases  yerrorlower=rr_l yerrorupper=rr_u  ;
refline 1/axis=y lineattrs=(pattern=dash);
colaxis  label=' '  ;
rowaxis  label=' ' ;
format Detroit $fmt.;
run;

Re: Plot RR (95% CI) paneled grid plots

Cruise — Mon, 22 Jun 2020 11:43:38 GMT

@Ksharp

Great! Thanks. I'm trying to bring bars closer.

Do you know how to keep decimals places to two consistent across the panels? Now, it ranges 1-3. keeping it to two or three is desirable.

proc format;
value $fmt
'yes'='All data'
'no' ='Detroit excluded';
run;
ods graphics / width=300px height=1000px;
proc sgpanel data=ci noautolegend;
styleattrs datacontrastcolors=(black red green);
panelby  Detroit Variables/ layout=lattice onepanel novarname proportional  
rowheaderpos=left noheaderborder colheaderpos=top  spacing=4 uniscale=column  ;
scatter x= Phases y=rr/ markerattrs=(symbol=squarefilled) group=Phases  yerrorlower=rr_l yerrorupper=rr_u  ;
refline 1/axis=y lineattrs=(pattern=dash);
colaxis  label=' '  ;
rowaxis  label=' ' ;
format Detroit $fmt.;
run;

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Mon, 22 Jun 2020 12:17:21 GMT

You could try

format _numeric_ 8.2 ;

put it in Data Step or Proc Report .

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Mon, 22 Jun 2020 12:29:11 GMT

OK.
Try this :

colaxis label=' ' ;
rowaxis label=' ' valuesformat=F8.2 ;
format Detroit $fmt.;
run;

Re: Plot RR (95% CI) paneled grid plots

Cruise — Mon, 22 Jun 2020 12:36:28 GMT

Hi ksharp, what will bring my bars closer to each other? spacing has no effect on it.

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Mon, 22 Jun 2020 12:48:15 GMT

OK. This one.

colaxis  label=' '  offsetmax=0.4 offsetmin=0.4  ;
rowaxis  label=' ' valuesformat=F8.2  ;
format Detroit $fmt.;
run;

Re: Plot RR (95% CI) paneled grid plots

Cruise — Mon, 22 Jun 2020 12:54:51 GMT

@Ksharp

Did the trick. Presto. 'Variables' in the ci data has long name in actuality. Do you know how to cut them to two lines to label on the y-axis? Right now, it is cut. After this, I'll post the final SAS code for the users who may have same plotting problem.

ods graphics / width=350px height=600px;
proc sgpanel data=ci_data1 noautolegend;
styleattrs datacontrastcolors=(black red green);
panelby  Daegu Variables/ layout=lattice onepanel novarname proportional  
rowheaderpos=left noheaderborder colheaderpos=top  spacing=2 uniscale=column;
scatter x= Phases y=rr/ markerattrs=(symbol=squarefilled) group=Phases  yerrorlower=rr_l yerrorupper=rr_u  ;
refline 1/axis=y lineattrs=(pattern=dash);
colaxis  display=none  label=' '  offsetmax=0.4 offsetmin=0.4  ;
rowaxis  label=' ' valuesformat=F8.2  ;
format Detroit $fmt. variables $variables.;
run;

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Wed, 24 Jun 2020 11:23:55 GMT

Sorry . I can't help you.

Post it at Graphic Forum

https://communities.sas.com/t5/Graphics-Programming/bd-p/sas_graph

@DanH_sas @GraphGuy maybe have answer .

Also calling @Rick_SAS @Reeza

Re: Plot RR (95% CI) paneled grid plots

Reeza — Wed, 24 Jun 2020 15:17:23 GMT

It's a spacing issue. You simply don't have enough space for the name in the graphic. There's an option to wrap the axis label using a split character and/or to rotate it to show different. I'd probably align it to the bottom (LABELPOS) and add in the split character. I'd suggest descriptive names using labels as well....

https://documentation.sas.com/?docsetId=grstatproc&docsetTarget=p0vyc21mth0jrtn1c4jimudemip4.htm&docsetVersion=9.4&locale=en

Re: Plot RR (95% CI) paneled grid plots

Ksharp — Thu, 25 Jun 2020 11:13:11 GMT

Reeza,
splitchar= fitpolicy= is for tick value ,not for y axis label. I tried but failed,
Even I tried "high_(*ESC*){unicode '000D'x} sch " to cut it as you pointed at another poster,but still nothing happen.

I thinked SAS would fix this problem ,since it looks like very useful for ODS Graphics.