Asked 22nd Feb, 2022

What is the next step after identifying the sample size for online survey?

I have found my sample size based on the population of the city by using online sample size calculators like Survey Monkey. However, the number I get is about 200-300 and I am wondering is this the number of surveys I should distribute, or should I multiply this number by a number (if yes, what number?) or something to account for non-respondant?
I have heard a 5-10% response rate is common, and 10% of 300 is not much... so I am a bit confused about how to proceed.
I appreciate any help!

All Answers (3)

23rd Feb, 2022
Rolf Frankenberger
University of Tuebingen
a sample size of 200-300 persons answering your survey seems to be ok if the city is not that big. Depending on how differentiated your analysis is intended to be (e.g. for living areas, districts and so on) you might want to have more participants.
A response rate of 5-10% is reasonable. So, you would have to contact about 3000 persons. If you want to do a online survey, this can be done by advertising on social media or by using some other way of dissemination, like e-mail lists.
If you do it by regular mail, response rate is also quite low.
If you want to do face to face, these might be less persons to be contacted, because the response rate is higher.
1 Recommendation
25th Feb, 2022
Viktor Bukovszki
Technische Universität München
Rolf is spot on, I would like to address a different aspect of your case. I do not know how SurveyMonkey computes sample size, but it is a good exercise to understand the statistical reasoning behind them. Take this source for example on page 103:
Now you have finite resources to contact people, and if you end up with a smaller sample that doesn't mean your results are not useful. This is why understanding the sample size formula matters, because at this point, you need to reverse engineer the satistical confidence in your results based on actual sample size. Also, statistical confidence does not equal confidence in general. You need to be transparent on the former, and try to explain the limitations and implications on the latter in your discussion. Just have this exercise of asking yourself, "in what ways can the data fool me?". It then depends on the risks associated with the decisions you inform whether the sample size need to be expanded or not.
1 Recommendation
2nd Mar, 2022
David L Morgan
Portland State University
That is the number of responses that you need in order to perform an adequate statistical analysis, so you definitely need to take the response rate into account.

Similar questions and discussions

Methodologically speaking, should the following situation be classified as a case study or is it better labeled as something else?
17 answers
  • Leônidas Soares PereiraLeônidas Soares Pereira
I have a question concerning how to classify this research approach. Let me give you the context:
A couple of colleagues and I found that a particular topic was showing itself to be particular tricky to teach in a classroom environment, thus we spent last year collecting literature on the topic and end up developing a theoretical framework for how to address the issue. We now plan to test our framework by implementing it in an undergrad classroom course that one of us is teaching.
Now here is the thing, we come from a social science background, and we intend to initially approach this effort qualitatively (through observation and semi structured interviews with participants). Thus, my initial reaction would be, from a methodological standpoint, to maybe call this a case study (see Yin, Merriam, Stake, etc.). However, considering that the person implementing the framework (and teaching the course) is also one of the researchers who developed it, can this still be called a case study? Or should it be considered something else? I've heard people suggest Action Research, Participatory Research, or event "a participant observer Case Study" but I'm not quite sure if this is quite it as well.
Any help would be appreciated, thank you.
Obs.: we are well aware of potential bias, and depending on the results of this first phase we then might, later on, run this again, this time as a more controlled experiment, comparing the framework version classroom outcomes’, against the ones from a "non-framework" one.

Related Publications

Got a technical question?
Get high-quality answers from experts.