I want to conduct an online survey for my bachelor thesis on sustainability in hospitality in NYC. The population is Generation Y. I'll use a convienience sample method. How do I justy the sample size. I was thinking of 200 just as a matter of feasibility and resources.
DEAR THOMAS GREETINGS
Because there is no magic bullet or single number, there are a few things you’ll want to have determined before you start figuring out what your sample size is:
Every questions of your survey is variable which are dependent on demographic factors. so you have to analyze all the variables.
Every question in your survey is a variable.
Sometimes surveys are tightly focused. "Who are you voting for in the election" and a few other questions might be included for interest, and you get results like 72% of people with an advanced degree will vote for Jane Doe in the next election.
Sometime surveys are for other purposes. Psychology surveys often as a range of questions, and the responses to these questions are combined into new variables that relate to specific emotions. It might be more effective to ask twenty questions related to Jane's performance in office, rather than bluntly asking "do you love or hate Jane's performance."
Sometimes the issue of interest is not simple. I want to know how farmer Bob deals with pesticide drift off his land. It might be useful to know if Bob knows about chemical trespass laws, if he uses windbreaks or various chemical or mechanical approaches. Is he aware of university fact-sheets, or relies on company fact-sheets.
An issue is risk. The larger the sample size the more likely it is that you will find at least one statistically significant result. The larger the sample size the more likely another researcher who tries to replicate your research will arrive at qualitatively the same answer. How do you feel about getting no significant results, or getting results that cannot be replicated?
For the survey, keep track of the response rate. If possible, try and some people who fail to respond the first time to give a response after a little coaxing. Keep track of these as a separate group. Or, on each survey place a rating on how much work it took you to get the individual to take the survey. If there is no difference in survey responses between the people that took the survey with little or no pressure and those that needed great encouragement, then it is more likely that the survey has broad applicability.
Overall, get as many surveys as you have time and resources to gather. This will give your results the greatest power. You should examine all significant results carefully to evaluate if the observed difference is large enough to be useful. Large sample sizes can overpower a study -- though I have yet to find a paper that argues "while these results were all statistically significant due to the large sample size, the effects were all so small that they can be ignored with no loss." The main problem with large sample sizes is that they provide an abundance of choices for how to conduct the data analysis. This can consume a great deal of time, with uncertain benefits to career development. If the "time and resources" criteria results in a value substantially less than the sample size in published works, then you will have to decide if your special circumstances warrant conducting the survey anyway or if you are better off doing something else. If the standard is 1000 and the best I can do is 200, can I convince reviewers that my hardships in gathering the data warrant publication?
Calculating sample size: I want to sample two populations with a 5 question survey. So 5*2*15=150 surveys. This will cause problems. Question 1: Male versus female.
Question 2: Income (5 levels).
Question 3: Race (4 choices).
Questions 4 and 5 are about rating the appearance of two chocolate cakes.
This experiment effectively has 2 populations * 5 incomes * 4 races as the treatments, or 40 treatments. A sample size of 150 is not even 4 surveys in each "treatment," and it is unlikely that the surveys will be equally distributed. You might have a great deal of information on Hispanic females making 75K/year, but no information at all on Native American males making over than 100K/year. This will cause trouble in the analysis. More surveys would help greatly, though you will always have the problem of unequal sample sizes unless you use purposive sampling methods. You may have to worry about non-uniform distribution of non-responses between each of the 40 groups.
Time: You can spend a great deal of time trying to estimate the ideal sample size before doing an experiment. The estimate is only as good as the guesses you used in doing the calculations, or the accuracy of the published values as they relate to the population of individuals that you are sampling. Spending this time gathering more surveys might (emphasis on might) be a better use of time especially if you know that you can get close to a publishable number of surveys.
Sample size is directly related to your interpretation. You have to justify sample size as you have to interpret the results. It needs some assumptions which you have to make at the time of planning a research like level of significance & power. If you specify 5% level of significance for sample size calculation, you will be only 95% confident to tell your result. So there is direct association. You have to calculate the sample size by assuming level of significance (LOS). You may calculate by taking 5% LOS and prevalence of the event which you are targeting. By doing this you may justify your sample size.
- Badges
- Science topic