Statisticians?

10 posts

Flag Post

Need some advice for my research:

I’m doing a study on a certain demographic group that purchases bulk ice cream (big ones, gallons) in supermarkets (go figure, can’t expect the street sellers to sell by the gallons).

I’m having some problems with my sampling method to use for obtaining respondents. I initially tried to use the convenience/accidental sampling, but I’ve consulted with a professor and she mentions trying to find a much stable sampling method (random sampling, etc), but they require a sampling frame, of which I’ve tried looking around for consumer groups that would fit the description of my target respondents, but to no avail. What then would be possible sampling methods to use that is stable enough, but doesn’t require any sampling frames.

 
Flag Post

It sounds like this would be a population without a sampling frame. Identify your population as precisely as possible. You say “certain demographic group that purchases bulk ice cream in supermarkets.” Is the demographic the group of people who purchase bulk ice cream in super markets, or is that only one of the characteristics of this group? You could do a phone or web survey with some screening questions about ice cream purchases and then get the information you need once you have a respondent that fits your criteria.

If you are doing a for-real-big-time study you could probably work something out to have the stores print an invitation to take a web survey on the receipts of the people who purchased gallons of ice cream. Many stores already do surveys like this for themselves, getting them to do it for you might take some extra work. Other than that your only real option is to send out a fleet of interviewers (with the permission of the stores, of course) to stake out the check out lanes and accost anyone who bought a gallon of ice cream.

Sometimes studies get done based on what information is available from whom rather than the original question that inspired the research :(

Good Luck, let us know what the answer is (or at least me, this is my kind of thing)

 
Flag Post

I’ve thought of all this already… I do have my target population’s characteristics, my survey questionnaire includes a screening process, etc. I’m just asking if I should stick to the convenience sampling method of obtaining data or should I use another one?

And it’s an academic thesis… quite small, and its basically a 1-man job (me).

 
Flag Post

Is there another sampling method that will get you respondents from your population? A probability method is preferred (it gets you generalizable results), but if you can’t get one you can’t get one and you take the convenience sample. Some data is better than no data.

Is this a masters thesis or some kind of senior project? I’m trying to figure out what you could be doing that you need data on large ice cream purchases for :)

 
Flag Post

It’s a senior project (or undergraduate studies). Doing a UAI study on certain consumers.

 
Flag Post

I’ve been looking unto Cluster Sampling, I may need to segregate it via locations, which I am partly doing. Do you think that may work?

 
Flag Post

To do a cluster sample in this situation (from what I understand of it) you would need to list all of the stores that sell ice cream by the gallon, and then choose x number of those at random. Then you will interview every single person you see buying ice cream by the gallon at those stores. In cluster sampling, your sample is the store, not the individual people buying ice cream in the store. To do this for real you will also need to do some hierarchal linear modeling with the numbers you get to avoid any bias because of the nested nature of your respondents. The people who buy their gallons at the fancy (expensive) store may have different purchasing habits (and attitudes) than the people buying gallons of ice cream at the store in the bad part of town.

X for your sample is based on the number of stores that sell gallons of ice cream and how accurate you want your results to be. The smaller the number of stores the larger your sample will need to be.

 
Flag Post

I’ve asked my adviser, she opted me to create another way of getting my respondents, I’ll think it over plus adding your inputs. I’ll let you know once I set things up.

 
Flag Post

Anyway, I’ve readjusted some things. I’ve decided to do the survey to parents who are attending PTA meetings in certain schools (topic & questionnaires still follows, just the locations and scenario got changed). Lists on potential visitors for the meetings would be my sampling base. I then initially chose the Random Sampling method, but being segregated into areas, I am also thinking of using the Cluster Sampling, but then, as schools, its not always on convenience or proximity, so it can work well in terms of random sampling, as students often come from different districts in the city.

 
Flag Post

I could use some big help here ASAP, hence the bumping…