topic Re: proc sgpanel x-axis with multiple scales in Graphics Programming

proc sgpanel y-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:11:48 GMT

I'd like use multiple scales on y-axis using below max min values.

agecat1 max= 12 min=3; agecat2 max=20 min=6, agecat3 max=20 min=5 agecat4 max=24 min=8; agecat5 max=27, min=7.

Could you please help modify the code below for multiple scales?

proc sort data=data; by race agecat cat year; run;
ods graphics / height=2000 width=3000;
proc sgpanel data=data;
panelby race agecat/novarname onepanel LAYOUT=panel columns=5 ROWHEADERPOS=right COLHEADERPOS=bottom;
series y=rate x=year / group=cat lineattrs=(pattern=solid);
keylegend/title="legend title" position=bottom;
colaxis label='column title' fitpolicy=thin valuesformat=best4.0 values=(2002 to 2018 by 2);
rowaxis label='rowtitle' grid max=27.0;
title "title";
run;

using SAS 9.4.

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 15:08:58 GMT

It's a bit hard to understand your question; but if you're asking how to make the X axis have different scales, you can use the option on the PANELBY statement UNISCALE which lets you choose between having all column axes the same but row axes different, or all row axes the same and column axes different (this is what you want), or both the same.

Re: proc sgpanel x-axis with multiple scales

PeterClemmensen — Mon, 08 Jan 2018 15:11:32 GMT

Don't you mean that you want different scales on your y-axes for different values of your ageat variable?

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:11:29 GMT

hey, yes. y-axis. i just edited my post. thanks!

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:13:45 GMT

My apology. I meant y-axis. I'm dealing with data that is different by age. so using same scale for all age categories is not making a perfect sense here. hope that makes sense.

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 15:17:44 GMT

That doesn't change my answer other than meaning you should choose UNISCALE=ROW instead of UNISCALE=COLUMN.

Re: proc sgpanel x-axis with multiple scales

PeterClemmensen — Mon, 08 Jan 2018 15:21:50 GMT

Ok. My first question is: why?

Usually you create a panel of graphs like this to be able to quickly compare characteristics of different group sin your data (in your example age and race). This comparability is worsened if you do not have the same values on your axes?

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:36:56 GMT

Plots on same scale should be kept for reference. Also, it's kids weight and height data where agecat is a modifier. So comparing newborn (agecat=1) to 5 year old kid (agecat=5) is not my primary research question but difference across race groups. Because of interactions, some graphs are collapsed and hard to eyeball,ie., middle three groups of agecat. Would you agree?

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 15:43:07 GMT

I think that the way you have it laid out above, I agree with Draycut; you should not change the column axes. What you should change is the entire presentation.

The reason to have a grid like this is to have the ability to compare left-to-right and up-to-down. So up and down makes sense: you have different races, and you want to be able to see the differences between asian and white or whatever. That's a good idea and makes perfect sense.

However, with the ages left to right, would you want to compare the newborns with the four year olds, by weight? I doubt it: you'd want to compare something *about* them, rather. So chart that. Or normalize your data, so instead of using actual weight/height, you normalize it to the 50th percentile for that age or something like that. But better I think would be to chart what you actually expect people to be looking at here, which I don't think is just the weight/height.

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:44:15 GMT

Snoopy:

Do you know why I get error below?

4984 ods graphics / height=1500 width=4000;

4985 proc sgpanel data=r.all_cnt;

4986 panelby race ageyears/novarname onepanel LAYOUT=panel columns=5 ROWHEADERPOS=right

4986! COLHEADERPOS=bottom uniscale=row;

4987 series y=pr_obs x=year / group=bf lineattrs=(pattern=solid) uniscale=row;

--------

ERROR 22-322: Syntax error, expecting one of the following: ;, ATTRID, BREAK, CLUSTERWIDTH,

CURVELABEL, CURVELABELATTRS, CURVELABELPOS, DATALABEL, DATALABELATTRS,

DATALABELPOS, DISCRETEOFFSET, FILLEDOUTLINEDMARKERS, GROUP, GROUPDISPLAY,

GROUPORDER, LEGENDLABEL, LINEATTRS, MARKERATTRS, MARKERFILLATTRS,

MARKEROUTLINEATTRS, MARKERS, NAME, NOMISSINGGROUP, SMOOTHCONNECT, SPLITCHAR,

SPLITCHARNODROP, SPLITJUSTIFY, TIP, TIPFORMAT, TIPLABEL, TRANSPARENCY, URL.

ERROR 76-322: Syntax error, statement will be ignored.

4988 keylegend/title="BREASTFEEDING DURATION" position=bottom;

4989 format race race. ageyears ageyears. bf bf.;

4990 colaxis label='COHORT YEARS' fitpolicy=thin valuesformat=best4.0 values=(2002 to 2015 by 1);

4991 rowaxis label='PREVALENCE OF OBESITY' grid max=27.0;

4992 title "OBESITY PREVALENCE BY";

4993 run;

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 15:45:23 GMT

You have UNISCALE=ROW on the SERIES statement, which is wrong.

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:49:51 GMT

Sorry for a confusion. This is kids' "weight and height for age percentile data" adjusted to gender. Where I defined obesity to 1 if kids percentile hits 97.7% else 0. And computed the obesity rate as: event count/n of kids in the strata (by age and race). I thought resulting data must be normalized since the source data was a normalized percentile data. isn't it?

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 15:55:06 GMT

And this is why it's helpful to give complete information in your question.

If you're normalizing it to age percentile, then why do you want to adjust the row axes? I think you should not alter the row axes for the different ages if it's adjusted by percentile: your chart is, correctly, showing that different age ranges have a different prevalence of obesity.

Re: proc sgpanel x-axis with multiple scales

Cruise — Mon, 08 Jan 2018 15:58:31 GMT

Thanks snoopy. Thanks for your insight!!! Off topic. Shall I create a separate post on how to output high resolution image out of this program?

Re: proc sgpanel x-axis with multiple scales

snoopy369 — Mon, 08 Jan 2018 16:01:43 GMT

Yes, if you want to ask something unrelated to the original question (as
that would be) please ask it in a different question. Thanks!

Re: proc sgpanel x-axis with multiple scales

ballardw — Tue, 09 Jan 2018 00:42:39 GMT

@Cruise wrote:
Thanks snoopy. Thanks for your insight!!! Off topic. Shall I create a separate post on how to output high resolution image out of this program?

You may want to define exactly what you mean by "high resolution image" as well.

Consider the final display size and memory requirements. If you have a "requirement" for 1000 dots per inch and a 10 inch by 10 inch display that is a lot of dots potentially and may exceed the program ability to manage.

Re: proc sgpanel x-axis with multiple scales

Yishi- — Tue, 14 Jun 2022 18:31:55 GMT

It works! Thanks