Please advise..

#1
I have collected 6 years of meeting registration data for about 400 meetings. I have the number of meeting registrants counted 24 weeks out from the date of the meeting. So for meeting abc I have something like this:

No. of weeks out No. of Registrants
6 3
5 6
4 11
3 17
2 22
1 20
0 24

Total Attendees: 103

Some meetings I have data for many weeks out for others the weeks out duration is a shorter-span (like the example above). My question is this:

If I have a meeting not yet started and it is say 3 weeks out from the meeting date. Can I predict what the final meeting attendee count is likely to be for the meeting 3 weeks away?

What type of analysis should I use? I am looking for my old college statistics textbook and have never felt so dumb. Thanks for any help or direction.
 

JohnM

TS Contributor
#2
For each meeting, compute the cumulative number of registrants for the number of weeks away, and divide that by the total number of final registrants to get a %. Then, for example, if you wanted to predict the final tally when the meeting is 3 weeks away, take all of your 3-week-out cumulative percentages and compute the average and standard deviation.

If, for example, on average, 38% of the people who attend have registered by the 3-week-out date, and there is a standard deviation of, lets say 5%, then a reasonable prediction band could be the mean +/- 2 standard deviations:

38% +/- (2 * 5%) = 28% - 48%

To predict the final attendance number, multiply the number representing 28% by 3.57 to get the lower end of the prediction band (28*3.57 = 100), and then multiply 48% by 2.08 (48*2.08 = 100) to get the upper end of the prediction band.
 
#3
Follow-up Question

I still need to ask one more question.

I have calculated my cumulative percent registered for each week out for each meeting.

I have the mean and the standard deviation for each week out.

I do not understand in the last paragraph of the response where the 3.57 or the 2.08 come from. How do I calculate these? And what is the ‘number representing 28%’ exactly? Is that the number registered at 3 weeks out for the meeting I am predicting?

Thanks again for any assistance.