I am going to conduct a true experiment design with pretest and posttest in the marketing field. The samples are randomly selected into the control group and experimental group. However, my questions if the samples should be representative? Since if yes, I need to ask a professional company and buy the samples; if not, I can utilize the snowball method to collect the samples and later on randomize them into two group.
Great thanks if someone experienced can give me some tips!
Samples must be representative of population. Snowball sampling technique is used in hidden populations which are difficult for researchers to access. As sample members are not selected from a sampling frame, snowball samples, are subject to bias.
Whatever the method of sampling you use that is diferent than representative one, will generate results not consistent to the population. If its not possible to implement the representative sampling you can treat samples as different populations and make conclusion on each of them.
Hello Mahfuz, hello Nikolay, thanks for the answers! I realized the method other than representative one will generate bias. The problem is that we do not have a database or sampling frame available to randomly distribute the questionnaire. As an alternative, I plan to conduct a mall intercept interview, i.e. randomly select the customers who visit the shop and ask their interest to participate the survey. How does it sound like?
yes you can do so You can refer to my paper too which followed the same approach
Impact of musical fits and image of different malls on consumer ...
www.inderscienceonline.com/doi/abs/10.1504/IJICBM.2016.075538
For experiments, random sampling is not that crucial while random assignment is. You can recruit the participants using a non-random sampling method, but for a TRUE experiment, the participants must be allocated to each treatment groups and the control group using random assignment. The crux is the equivalency of the groups rather than their representativeness. A more heterogeneous sample may create the need for matching plus random assignment to achieve equivalency. I hope this helps.
It is very much depending on the experiment that you are doing. I fully agree with Imam that random sampling is not always crucial. The technique that you are planning to follow is valid and we have used it many times when we do not have a valid database. Be aware of marketing research agencies data their focus is to have sufficient data to accommodate their customers needs and not to be scientifically correct.
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology