topic Re: Flag overlapping 95% CLs in SAS Programming

Flag overlapping 95% CLs

SAS93 — Mon, 18 Jul 2022 15:05:35 GMT

I want to find a way to flag or otherwise ID the ranges of 95% CLs for a series of datasets I have.

Obs	FPL	RowPercent	RowLowerCL	RowUpperCL
1	1	67.4	63.5	71.4
2	2	68.0	62.5	73.6
3	3	63.8	56.2	71.4
4	4	51.1	42.0	60.3

These datasets are outputs from previous crosstabulations. I run Chi-square tests for these crosstabulations as well, but I want a visual way to ID and get a sense of which CLs overlap with all the others (not just compare FPL=1-2, 2-3, etc.).

I've come across various approaches to ID consecutive data ranges that overlap, largely with first./last. and retain statements, but they don't work for my situation. I think a matrix might be best to concisely ID the unique pairs, but I'm not sure how to approach that.

Re: Flag overlapping 95% CLs

PaigeMiller — Mon, 18 Jul 2022 15:10:57 GMT

Please show us the code that created these confidence intervals.

Re: Flag overlapping 95% CLs

ballardw — Mon, 18 Jul 2022 15:16:14 GMT

Something like this perhaps:

data have;
   input FPL 	RowPercent 	RowLowerCL 	RowUpperCL;
datalines;
1 	67.4 	63.5 	71.4
2 	68.0 	62.5 	73.6
3 	63.8 	56.2 	71.4
4 	51.1 	42.0 	60.3
;

proc sgplot data=have;
   highlow x=fpl  low=rowlowercl high=rowuppercl;
   xaxis type=discrete;
run;

Note the use of data step code to provide example data. That way we can write code to use your example data.

Re: Flag overlapping 95% CLs

Rick_SAS — Mon, 18 Jul 2022 17:47:21 GMT

If you are trying to see whether the means of different categories are significant;y different from each other, be sure to correct for multiple comparisons. Please see the visualizations that are created automatically by using PROC GLM, such as

Be aware that the overlapping of 95% CIs is neither necessary nor sufficient for inferring that the difference between the means is significant.