Time - sampling non-independent sample (1 viewing) (1) Guest
Favoured: 0
|
|
|
TOPIC: Time - sampling non-independent sample
|
|
|
|
Time - sampling non-independent sample 13 Years, 10 Months ago
|
|
|
Hello,
I am looking to do a quant study on market size share in the tourism industry. What we would like to do is to sample a random group of 1000 respondents each month (representative for the standard sociodemo's in my country), to probe for their leisure activities in that month (such as going to the museum, to the zoo, etc.).
However, one could stongly argue that observations are not independent between months: what I will do in a certain month (for instance, visit a particular ZOO) is dependent on what I have done the months before. If I did visit a particular ZOO last month, chances are lower that I will visit that Zoo again this month.
Given this "non-indepency", would you argue to use the same set of 1000 people for every of the 12 waves?
Or would you say that having a new random (but representative) sample evey month would suffice?
Drawback of the first approach is that it's three to four times as expensive, since you have to recruit much 3000-4000 people to end up with 1000 at the end.
I'm trying to get my head around it but I can't find the correct answer. Intuitively I would opt for the "one fixed sample" aproach, but thinking about the "12 samples" I cannot pinpoint the exact drawbacks. There are going to be influence effects from one month to the other, but I would think that the same effects would be there in both approaches.
What do you think? WHich is the best method, and why?
|
|
|
|
|
|
|
The administrator has disabled public write access.
|
|
| |
|
 |
|
dejohan
|
2012/10/08 17:14
|
|
|
|
|
|
Forum 
