topic Re: proc sgplot histogram producing only a single histogram bar in SAS Enterprise Guide

proc sgplot histogram producing only a single histogram bar

RobertWF1 — Tue, 04 Jan 2022 17:49:18 GMT

I'm trying to create a histogram with proc sgplot but SAS will only create a single bar histogram plot.

My data is continuous, highly skewed, with 10,677 zero values. When I run tests on simulated lognormal data, proc sgplot does fine.

Perhaps something about my data is causing issues?

proc means data=out.cohort_match_real_data_pos n mean median stddev min max; where ever_eng_fst12_month="No"; var pre_post_cost_712_mon_ratio; run; /* N Mean Median Std Dev Minimum Maximum */ /* ----------------------------------------------------------------------------------------*/ /* 37670 8.9096483 0.4685912 117.5225125 0 15089.56 */ /* ----------------------------------------------------------------------------------------*/ proc sgplot data=out.cohort_match_real_data_pos; where ever_eng_fst12_month="No"; histogram pre_post_cost_712_mon_ratio / nbins=40 binwidth=.5 binstart=0; /* density pre_post_cost_712_mon_ratio /type=normal;*/ xaxis values=(0 to 20 by .5); run;

Re: proc sgplot histogram producing only a single histogram bar

PaigeMiller — Tue, 04 Jan 2022 17:48:31 GMT

Are there messages in the log that might indicate what the problem is?

From now on, when code doesn't work, show us the log (all of it for this PROC, not just the ERRORs or WARNINGs)

Re: proc sgplot histogram producing only a single histogram bar

Reeza — Tue, 04 Jan 2022 17:52:13 GMT

Your maximum value is 15089 but your X axis only goes to 20?

@RobertWF1 wrote:

I'm trying to create a histogram with proc sgplot but SAS will only create a single bar histogram plot.

My data is continuous, highly skewed, with 10,677 zero values. When I run tests on simulated lognormal data, proc sgplot does fine.

Perhaps something about my data is causing issues?

proc means data=out.cohort_match_real_data_pos n mean median stddev min max; where ever_eng_fst12_month="No"; var pre_post_cost_712_mon_ratio; run; /* N Mean Median Std Dev Minimum Maximum */ /* ----------------------------------------------------------------------------------------*/ /* 37670 8.9096483 0.4685912 117.5225125 0 15089.56 */ /* ----------------------------------------------------------------------------------------*/ proc sgplot data=out.cohort_match_real_data_pos; where ever_eng_fst12_month="No"; histogram pre_post_cost_712_mon_ratio / nbins=40 binwidth=.5 binstart=0; /* density pre_post_cost_712_mon_ratio /type=normal;*/ xaxis values=(0 to 20 by .5); run;

Run the code without any options and add them back in one by one to see where the issue is.

proc sgplot data=out.cohort_match_real_data_pos;
	where ever_eng_fst12_month="No";
 	histogram pre_post_cost_712_mon_ratio ;
run;

Re: proc sgplot histogram producing only a single histogram bar

RobertWF1 — Tue, 04 Jan 2022 17:55:41 GMT

Here's the log - no errors:

23 GOPTIONS ACCESSIBLE;
24 proc sgplot data=out.cohort_match_real_data_pos;
25 where ever_eng_fst12_month="No";
26 histogram pre_post_cost_712_mon_ratio / nbins=40 binwidth=.5 binstart=0;
27 /* density pre_post_cost_712_mon_ratio /type=normal;*/
28 xaxis values=(0 to 20 by .5);
29 run;

NOTE: PROCEDURE SGPLOT used (Total process time):
real time 7.51 seconds
cpu time 5.06 seconds

NOTE: The specified BINWIDTH= value will be ignored in order to accommodate the data.
NOTE: Some of the tick values have been thinned.
NOTE: Listing image output written to D:\SASWORK\rfeyerharm\_TD3252_VMSAS01_\Prc2\SGPlot38.png.
NOTE: Some of the tick values have been thinned.
NOTE: There were 41355 observations read from the data set OUT.COHORT_MATCH_REAL_DATA_POS.
WHERE ever_eng_fst12_month='No';

Re: proc sgplot histogram producing only a single histogram bar

PaigeMiller — Tue, 04 Jan 2022 18:00:25 GMT

It doesn't have to be ERROR message, the answer is right here if you would just read the log: (Maxim 2, Read the Log)

NOTE: The specified BINWIDTH= value will be ignored in order to accommodate the data.

So SAS chooses its own binwidth, which apparently is very wide. If you only want data between 0 and 20 to appear in the histogram, eliminate any values >20

Re: proc sgplot histogram producing only a single histogram bar

RobertWF1 — Tue, 04 Jan 2022 18:03:37 GMT

If I run sgplot with no options, I get the following default plot (same with proc univariate histogram).

It's lumping most of my observations into one category.

Re: proc sgplot histogram producing only a single histogram bar

Reeza — Tue, 04 Jan 2022 18:09:43 GMT

Yes, because a large proportion are zeroes as indicated.
Why not exclude those to see the distribution of the remainder?

where ever_eng_fst12_month="No" and pre_post_cost_712_mon_ratio > 0 ;

Re: proc sgplot histogram producing only a single histogram bar

RobertWF1 — Tue, 04 Jan 2022 18:11:01 GMT

That worked!

proc sgplot data=out.cohort_match_real_data_pos;
where ever_eng_fst12_month="No" and pre_post_cost_712_mon_ratio <= 20;
histogram pre_post_cost_712_mon_ratio / nbins=40 binwidth=.5 binstart=0;
/* density pre_post_cost_712_mon_ratio /type=normal;*/
xaxis values=(0 to 20 by .5);
run;

Re: proc sgplot histogram producing only a single histogram bar

RobertWF1 — Wed, 05 Jan 2022 15:20:49 GMT

Follow up question:

Is there a way to change my x-axis in proc sgplot without changing the percentages for each bin? I understand SAS is recalculating the %'s when I subset the population, which it's supposed to do.

But if I want to "zoom in" on a portion of the x-axis without renormalizing the %'s, is there a handy option in sgplot? Or better to manually calculate the bins and %'s, say with proc freq & output a bin_freq dataset, and then run sgplot on bin_freq.