topic Re: PROC GCHART graphic area problem in Graphics Programming

PROC GCHART graphic area problem

AncaTilea — Tue, 12 Feb 2013 15:16:50 GMT

Hi.

I have a working code that manifests differently for what seems to me very similar conditions. PLease see below:

I have a data set that has estimates from a couple of models (one assessing within the other one between subjects...)

I initially had a chart that plotted the estimates from both models in one plot:

data have;

input measure $ 1-19 model $ Estimate ;

cards;

First, Measure b_gfr 3.48

First, Measure d_gfr_10 3.56

Second, Measure b_gfr 1.38

Second, Measure d_gfr_10 2.22

Third, Measure b_gfr 1.40

Third, Measure d_gfr_10 2.03

Fourth, Measure b_gfr 1.38

Fourth, Measure d_gfr_10 3.28

Fifth, Measure b_gfr 0.85

Fifth, Measure d_gfr_10 1.51

Sixth, Measure b_gfr 0.07

Sixth, Measure d_gfr_10 0.38

Seventh, Measure b_gfr 0.63

Seventh, Measure d_gfr_10 0.93

Eighth, Measure b_gfr 2.99

Eighth, Measure d_gfr_10 5.17

;

filename models "&plot_path.\models_both.jpeg";

goptions reset = all device = jpeg gsfname = models gsfmode = replace ;

options orientation=landscape nodate nonumber ;

pattern1 color = graycc;

pattern2 color = gray;

axis1 label = none minor = none order = (0.0 to 7.0 by 1) value = none major = none c = white;

axis2 label = none value = none;

axis3 split = "," label = none value = (f = "Arial/Bold" h = 1.2 ) c = black;

legend label = none position = (inside top left) mode = protect across = 1

value = (h = 1.7 f = "Arial" "Between-Patient Models" "Within-Patient Models");

proc gchart data = have ;

vbar model/

sumvar = estimate NoZero space = 0

group = measure subgroup = model legend = legend

type = mean iframe = black

raxis = axis1 maxis = axis2 gaxis = axis3

coutline = black

;

run;quit;

It turns out the request is now for separate plots. Very easy to do, I just split the code and restrict the data to one mode at the time, using a where statement.

filename models "&plot_path.\models_cross.jpeg";

goptions device = jpeg gsfname = models gsfmode = replace ;

options orientation=landscape nodate nonumber ;

axis1 label = none minor = none order = (0 to 7 by 1)

value = none major = none c = white;

axis2 label = none value = none;

axis3 split = "," label = none value = (f = "Arial/Bold" h = 1.2 ) c = black;

pattern color = grayee;

legend label = none position = (inside top left) mode = protect across = 1

value = (h = 1.7 f = "Arial" "Between-Patient Models");

proc gchart data = have ;where model = 'b_gfr';

vbar model/

sumvar = estimate NoZero space = 0

group = measure legend = legend

type = mean iframe = black

raxis = axis1 maxis = axis2 gaxis = axis3

coutline = black

;

run;quit;

filename models "&plot_path.\models_within.jpeg";

goptions device = jpeg gsfname = models gsfmode = replace ;

options orientation=landscape nodate nonumber ;

axis1 label = none minor = none order = (0 to 7 by 1)

value = none major = none c = white;

axis2 label = none value = none;

axis3 split = "," label = none value = (f = "Arial/Bold" h = 1.2 ) c = black;

pattern color = gray;

legend label = none position = (inside top left) mode = protect across = 1

value = (h = 1.7 f = "Arial" "Within-Patient Models");

proc gchart data = have ;where model = 'd_gfr_10';

vbar model/

sumvar = estimate NoZero space = 0

group = measure legend = legend

type = mean iframe = black

raxis = axis1 maxis = axis2 gaxis = axis3

coutline = black

;

run;quit;

OK, so my problem is that one plot 'fills' the graphic area, the other one seems to be smaller and I can't figure out why.

Any help would be highly appreciated.

Thank you!

Anca.

Re: PROC GCHART graphic area problem

GraphGuy — Tue, 12 Feb 2013 16:41:11 GMT

Your code seems a bit overly-tricky, using the group= option for what is visually not a grouped bar chart.

I would use the following simplified code instead:

axis1 label=none minor=none order=(0 to 7 by 1) value=none major=none c=white;

axis2 split="," label=none;

pattern color=grayee;

proc gchart data=have; where model='b_gfr';

vbar measure /

sumvar=estimate NoZero

legend=legend

type=mean iframe=black

raxis=axis1 maxis=axis2

coutline=black;

run;

pattern color=gray;

proc gchart data=have; where model = 'd_gfr_10';

vbar measure /

sumvar=estimate NoZero

legend=legend

type=mean iframe=black

raxis=axis1 maxis=axis2

coutline=black;

run;

Re: PROC GCHART graphic area problem

AncaTilea — Tue, 12 Feb 2013 16:58:41 GMT

Thank you!

:smileygrin:

Re: PROC GCHART graphic area problem

GraphGuy — Tue, 12 Feb 2013 19:11:32 GMT

I would also recommend using device=png (or device=gif), rather than device=jpeg.

Re: PROC GCHART graphic area problem

AncaTilea — Tue, 12 Feb 2013 19:14:40 GMT

I was wondering why users prefer the .png to .jpeg?

is there any reason for that?

My graphs tend to look acceptable in jpeg, and eventually when time gets to publish I make .eps or .tiff...

If you have more thoughts on this, I would appreciate your opinion.

Thank you in advance.

Re: PROC GCHART graphic area problem

GraphGuy — Tue, 12 Feb 2013 19:26:39 GMT

The jpeg format is more geared towards photographs, and can produce "blurry" looking graphs & charts (or at least that's been my experience).

The png or gif formats are better for creating "crisp" looking graphs & charts. Also, png supports anti-aliasing (of text & lines) and supports millions of colors. I recommend using png.

Re: PROC GCHART graphic area problem

GraphGuy — Tue, 12 Feb 2013 19:35:32 GMT

Or, in the words of our developer who works on the SAS jpeg driver ...

"The JPEG format is by nature more suited to reproducing photographic images than

business graphics due to the lossy compression method used. PNG, TIFF, and GIF

are better format choices for plots, charts, and other business graphic

drawings. It is possible to clean up JPEG images a little by increasing the image

quality setting to 100 per cent (using the new jpegquality system option)."