Graphs, Funnel Plots, suggestions?

[jamesmartinn]

Hi!

I'm trying to display some data, let me describe it to you.

I have a group of about 5000 teachers. For each teacher, I have the rate of passing students (calculated by the total number of passes / total number trials) over a years period. For each teacher, I have the associated specialties (teachables): Math, Biological sciences, social sciences, etc. Lastly for each teacher, I have the amount of students they've had over a 3 year period (we call this quantity "Volume"). What I would like to show is the relationship between Passing Rate and Volume, by Teaching Type.

It was suggested I try a funnel plot. Does this seem appropriate? Any other suggestions? Thanks

 

[hlsmith]

Perhaps a forest plot may be more appropriate for presentaton of final data. Funnel plots typically get at random effects. But you may do both. Are you thinking odds ratios?

 

[jamesmartinn]

Perhaps a forest plot may be more appropriate for presentaton of final data. Funnel plots typically get at random effects. But you may do both. Are you thinking odds ratios?

This is kind of where i'm confused.

In some of the funnel plots I've seen, they've used counts, standardized rates or some other metric.

Here I have percentages (#events/#total trials) as the Y axis and I'd like Teacher Volume as the X axis.

I kind of want to show that there might be a relationship between how much volume a teacher has had and their passing rate and that this relationship could be a function of teaching type. Teaching type is highly variable in the data, with one specific teaching type having a lot more volume than the rest (math teachers, since in the context of the research, all students are required to take math)


I've been looking at this site: http://blogs.sas.com/content/iml/2011/11/23/funnel-plots-for-proportions/

I'm not sure how to adapt it to my data. Ideally, I'd like the confidence limits be for the overall proportion, aggregated (which is 43% disregarding teacher type).


Edit: In the website, they have plotted the # trials (as the population values I suppose) on the X axis. Here I want to add Teacher Volume.

any further advice? thanks for the reply!

 

[Dason]

Did you try just making scatterplots of passing rate by volume for each of the teaching types?

 

[jamesmartinn]

Thanks for the reply, Dason. This is exactly what I wanted to do (and did, though off to the side). The PI is insistent on the idea of Funnel Plots but the more I look into it, I'm just not sure if it's necessary. But yes, basically it would be an X-Y scatter, by Teaching Type I agree.