I am trying to study the technology adoption and customer satisfaction with respect to mobile phone users in India. So it makes the total population for my study as 400 Million mobile phone users in the whole country. In this scenario how shall I proceed to collect the data (technique) and what shall be the adequate sample. Given the scale has over 70 items.
There are several methods. Random Digit Dialing (RDD) is the best known approach. Häder et al. (eds., 2012) give an overview about telephone surveys including mobile phones in Europe:
http://www.springer.com/us/book/9783642254109
Also see the AAPOR cell phone task force report for a good overview on RDD:
https://www.aapor.org/AAPORKentico/Education-Resources/Reports/Cell-Phone-Task-Force-Report.aspx
The ideal framework presupposes you would have a list of all the mobile users and the mean to contact with them. If you do not have this list, it is practically impossible to select a probabilistic sample from which infer. But if you have this list, you could select a systematic sample of it. For example one user (the random first user of the group) from a million of them. Then you will have 400 users in the sample. If you select one user from each 5 millions of users, you will have a sample of 80 users. The sample mean (of 80 users) estimates without bias the population mean (of 400 millions of users). The variability of the estimation can be obtained from the same sample.
The theory is in:
Cochran, W. G. (1977). Sampling Techniques, 3rd edn. Wiley. New York. [For mathematicians].
Thompson, S. K. (2012). Sampling, 3rd edn. Wiley. Hoboken, NJ. [For technicians].
The sample size depends of the required precision or confidence level, and of the randomization used. With systematic sampling of single start, this is impossible of estimating since it is not possible to estimate unbiasedly the population variance. But with double start, yes, it is possible but not easy. It requires a sampling expert. A simple random sample I think it is better (with equal probabilities of selection in each drawn of mobile user of the sample). With two phases, (1) pilot, and (2) study, I recommend to estimate the population variance in the pilot sample, and complete the remaining units of the sample with the estimated sample size in the first phase (pilot).
Can you share the scale developed by you for your study. Revert back at [email protected]
- Badges
- Science topic
- Similar topics
- Statistics
- Survey