BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
DarthPathos
Lapis Lazuli | Level 10
Hi all, I've used Rick Wicklin's post on creating funnel plots (http://blogs.sas.com/content/iml/2011/11/23/funnel-plots-for-proportions.html) successfully on my data. The challenge I'm facing now is figuring out how to label the outliers (above and below) based on either the 95% or 99.5% levels. I can apply labels when the threshold is a set number on the X or Y axis, but I can't seem to find anything on how to do it based on the proportion. Thanks for your time Chris Here's an example of what I'm looking to do, taken from https://blog.ouseful.info/2011/10/31/power-tools-for-aspiring-data-journalists-r/. The article does not provide the code or math to accomplish this, rather it's how to build a simplistic version in R (which doesn't help me).
bowel-cancer-mortality-ra-007.jpg
Has my article or post helped? Please mark as Solution or Like the article!
1 ACCEPTED SOLUTION

Accepted Solutions
Rick_SAS
SAS Super FREQ

You need to merge the data and the limits. To help do that, use all integer values for the limits. You need to make changes in three places:

 

1) In the IML program, there is a definition of 'n' as

n = T( do(minN, maxN, (maxN-minN)/20) );

Replace that line with all possible trials:
n = T(minN:maxN);

 

2) Merge the data and the computed limits:

/* merge data with proportions */
data Adoptions; merge Adoptions Stats; run;

/* merge control limits */
proc sort data=Adoptions; by Trials; run;

data Funnel; 
   merge Adoptions(rename=(Trials=N)) Limits; 
   by N;
   Label = Authority;
   if L3sd <= Proportion <= U3sd then Label=""; /* delete if w/in +/- 3*sigma */
run;

3) Change the SCATTER statement to 

scatter x=N y=Proportion / datalabel=Label;

 

 


SGPlot.png

View solution in original post

6 REPLIES 6
Rick_SAS
SAS Super FREQ

For each observation, you can tell whether it is outside of the 95% confidence interval. Create a label variable that is missing for points inside the CI and .has nonmissing values for points outside the CI, as shown in the article "Label only certain observations with PROC SGPLOT."

DarthPathos
Lapis Lazuli | Level 10

Hi @Rick_SAS

Thanks for the reply - I actually spent quite a bit of time yesterday working thourhg that blog, as I figured it was the easiest thing to do.  Where I'm stumped is in the "Funnel" dataset, we have "Authority", "Events", Trials", "Proportion", and "Expected"; then, rows 144 - 164 (in the same dataset) have "N", "L3sd", "L2sd", "U2sd", and "U3sd".  If I understand correctly, what I need to do is match the number of events for the individual authority (for example, Halton has n=20) with the closest N (in my case, probably N=26) so that i can get the lower/upper CI of 0.435 and 0.956.  Where I'm getting stuck is figuring out how to do this; I tried splitting the tables into 2 and then doing a join in SQL but couldn't get it to work.

 

Based on the article you linked to, here's my result based on Proportion <=0.5 and Proportion >=0.95.  This works fine for the N<=60 or so, but because the Proportions drop (and therefore the CIs get narrower), this does not work for the right side of the graph.  What I'd like is all the points above the upper red line and below the lower red line to be labelled.

 

Thanks again for your help.  I'm hoping I have explained this clearly - I'm multi-tasking (trying to do work and this at the same time :-)).

 

Chris


image001.png
Has my article or post helped? Please mark as Solution or Like the article!
Rick_SAS
SAS Super FREQ

You need to merge the data and the limits. To help do that, use all integer values for the limits. You need to make changes in three places:

 

1) In the IML program, there is a definition of 'n' as

n = T( do(minN, maxN, (maxN-minN)/20) );

Replace that line with all possible trials:
n = T(minN:maxN);

 

2) Merge the data and the computed limits:

/* merge data with proportions */
data Adoptions; merge Adoptions Stats; run;

/* merge control limits */
proc sort data=Adoptions; by Trials; run;

data Funnel; 
   merge Adoptions(rename=(Trials=N)) Limits; 
   by N;
   Label = Authority;
   if L3sd <= Proportion <= U3sd then Label=""; /* delete if w/in +/- 3*sigma */
run;

3) Change the SCATTER statement to 

scatter x=N y=Proportion / datalabel=Label;

 

 


SGPlot.png
DarthPathos
Lapis Lazuli | Level 10

thanks Rick - that was perfect and I actually think I understand what you did 🙂

 

Have a great evening

Chris

Has my article or post helped? Please mark as Solution or Like the article!
Rick_SAS
SAS Super FREQ

Although simple, tt's a bit of overkill to have 'n' contain all integers. The important thing is that it contains all the trial values in the data, as well as enough evenly spaced values to visualize the funnel.  Then you can do the merge, rather than interpolate.

DarthPathos
Lapis Lazuli | Level 10

thanks for the further explanation.  definitely one of the cooler analyses I've worked on in a while!

 

Chris

Has my article or post helped? Please mark as Solution or Like the article!

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

How to Concatenate Values

Learn how use the CAT functions in SAS to join values from multiple variables into a single value.

Find more tutorials on the SAS Users YouTube channel.

Click image to register for webinarClick image to register for webinar

Classroom Training Available!

Select SAS Training centers are offering in-person courses. View upcoming courses for:

View all other training opportunities.

Discussion stats
  • 6 replies
  • 1986 views
  • 3 likes
  • 2 in conversation