Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite

I have over 200 datapoints in my set and I have related my predictors (mostly categorical) to a single continuous response variable in a multiple regression model. The model is to my satisfaction, and now I need help in analyzing one aspect of it.

In my case, each datapoint represents an event, and my response variable is the number of attendees for that event. Increasing event frequency seems to decrease the number of attendees per event. However, so far increases in frequency have increased the OVERALL event attendance. So what I'd like to do is project the number of attendees for an entire week given various numbers for the total event count that week. Basically, I want to make a table with a column for the number of events that week, and another for number of total attendees for that week in order to select the optimal event frequency.

I have a variable in my model for number of events held that week. I know that I can use its coefficient to determine the negative impact of holding one additional event in a given week, but that data isn't useful to me without a starting point to subtract it from. I'm looking for a way to predict the number of attendees for a given weekly frequency ignoring all other variables. Is there some kind of weighted average that will help me find this number?

share|improve this question
I want to make sure I understand this part correctly : > Increasing event frequency seems to decrease the number of attendees per event. However, so far increases in frequency have increased the OVERALL event attendance. Do you mean that : - if you do a simple linear regression of the number of attendees on the event frequency only, you get a positive effect, whereas - in the multiple regression model, the effect of frequency is negative ? – Karim L Mar 5 at 22:47

1 Answer

up vote 0 down vote

You could use the mean (or median) values for all other variables and then vary the one variable of interest to you.

share|improve this answer
The problem here is that most of my other variables are categorical, and with many levels. How can I plug in a mean for those? – Pacific 231 Aug 7 '12 at 15:48
Can you give an example of one of the categorical variables that has many levels? – Joel W. Aug 7 '12 at 17:06
Sure. One of them is the name of the speaker hosting the event. We draw from a pool of at least 50 such speakers... – Pacific 231 Aug 7 '12 at 18:03
Sorry, but I do not understand your situation. How do you enter this speaker-hosting variable in the regression model? Do you have 50 yes-no variables, one for each speaker? How many data points do you have, that you based your multiple regression model on? – Joel W. Aug 7 '12 at 18:56
1  
You may have too many predictors for your 200 data points. Your multiple R may be large, but the large value may be an artifact of the relatively large number of predictors you are using. (Look up "shrinkage" or "capitalizing on chance" in any reference on multiple regression for more details.) How many predictors do you have in all? – Joel W. Aug 9 '12 at 12:38
show 2 more comments

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.