A typical dissertation would need you to include primary data. Primary data includes results that you obtain from a survey. A survey would contain questions, where you ask respondents to answer them. You would then need to analyse the results in order to prove or disprove your hypothesis or to answer the problem of your assignment. This seminar is to explain to everyone the best way of creating and conducting survey.
A good dissertation would have a clear objective and problem to be solved. We have produced a short article regarding planning and writing your dissertation as well as a research proposal. Once this is done, you should be able to identify the target population.
When trying to obtain the data, you need to know where you get the data. When the data is from the survey, you will need a group of people (also known as sample) to answer the questions. These people are known as the target population.
Back to Top
Choosing the correct population
You must now decide on the right people to ask and location is important as well. This is essential otherwise your response will be biased, thus your analysis would be deemed invalid by the tutor. This can also lead to losing unnecessary marks in your project. An example of a biased response is when you attempt to find out whether respondents prefer McDonalds over KFC by conducting the questionnaire outside a McDonalds’ restaurant. Of course the people you meet outside a McDonalds are extremely more likely to prefer McDonalds.
This may sound basic, but you must picture what type of results you expect to come back. By this, I don’t mean to ask you to predict the amount of answers in each question. You must know whatever the responses that will come back, would you be able to use the results to do the necessary hypotheses testing if required. Would it be relevant in solving your research problem?
Back to Top
Choosing the sample size
You need to determine the appropriate margin of error. What is margin of error? Well, it is the reliability of the survey results. If you are trying to generalise the opinion of a population, then your result would be more accurate if you ask more people. For example, if you ask 10 people within a large population, and 6 people say they travel to work using cars, then it will be highly inaccurate to say 60% of population travel to work using cars. So how do we make sure that the sample size is big enough to ensure accuracy?
Let us have a look at the simple calculation below. The formula to calculate this is 1/√N. N is the number of surveys. The chart below shows the sample size and margin of errors
What does the number mean above? Let us suppose you have decided to ask 10 people in your survey.
In one question, there are 6 respondents chose answer(a) and 4 respondents chose answer(b). The margin of error is 31.62% or 32% for 10 respondents. This means that the actual number that chose answer (a) could be as low as 28% (60-32) or as high as 92% (60+32). This range of margin error is too large. We have designed a margin of error calculation in Excel for you to use.
While choosing a large sample size is better, you need to take time and cost into consideration. For example, let us have a look at the margin of error table again.
As we can see, the margin of error for 1000 participants is 3.16% and for 2000 participants, is 2.24%. You need to determine whether it is worth it to undertake 2000 participants just for a 1% change in the margin of error. Remember that you also have limited time during your dissertation!
Our questionnaire should have a title in order to inform respondents what sort of questionnaire they are undertaking. There should be a short welcome message telling the respondents what the questionnaire is about, how long it will take to complete the questionnaire and confidentiality information. Be brief and keep it short and simple.
Thank you for taking part in this questionnaire. The nature of this questionnaire is to study XXX. Your response will provide significant contribution to the academic research at our University of XXX. Your personal information will be kept confidential and all responses will remain and be kept anonymous in future publications. Please read each question carefully before making a response. This questionnaire should take about XX minutes to complete. Thank you in advance for taking the time to fill in this questionnaire.
Hi, I am *name from *name of university. Would you please help me with my research by completing this short survey? If the respondent agrees to complete your questionnaire, you can use the same welcome message as above.
You may have a lot of knowledge on technical words because you have been studying the subject. If you are asking experts in your field to complete the survey, it is okay to use a lot of technical words. However, if you are asking an average person who may not know much about your subject area, you should keep the wording simple.
Make sure the questions are relevant and to the point. You do not want too many questions as the respondents will not be too happy if they spend too much time completing it. At the same time, too little questions will make your research incomplete or you may risk a double barrelled question. Keep it to the point and make every question relevant to your research. A good place to start will be not asking too many personal questions if they are not useful to your research such as name, age, height, hair colour etc.
Back to Top
Questions to avoid in a survey
Biased question Results can be biased when you ask an influential question. Biased questions are questions, in which you have already influences the respondents as to a certain answer.
It is common knowledge that McDonald serves high calories food. Do you think this is the case?
A smart person would say TV is a bad influence for children. Do you think this is the case?
How would you rate the calories content in McDonald’s food?
What is your opinion regarding TV’s influence on children?
When you give out what appears to be fact, you already influence the person that responds to it. Also, for the second example, it is sort of telling the respondent that they are stupid if they disagree with “TV is a bad influence for children”. The last problem with both examples, is that although they do not appear biased, is that the answer is either a “Yes” or “No”. This is fine if that is the answer that you want but it may limit the amount of responses for you to conduct any quantitative test or even qualitative analyses.
Another mistake that could happen is when you ask a question that is “double barrelled” or contains two statements. Basically you are asking two issues within a question and ask the respondents to accept or disagree with both issues at the same time.
Please indicate if you agree or disagree with the following statements:
I prefer to use bus and train to get to work
British Airways and Virgin Atlantic offer the best customer service.
As you can see, the examples above have two issues going on at the same time.
For the first example, what if you prefer to use the train but hate the bus? Same for the second question, what if you had a negative experience with Virgin Atlantic and a positive experience with British Airways?
The best way to ask such questions is to split each question into two. Corrections: I prefer to use the bus to get to work I prefer to use the train to get to work British Airways offers the best customer service Virgin Atlantic offers the best customer service
A double negative questions is when you attempt to ask someone to disagree with negative statements. The confusion is not just for the respondents, but you could also easily make a mistake. This is because, if you disagree with a negative statement, you essentially agree to the overall scenario.
It is impossible for the London Underground not to provide a good service for its customers
Do you feel that alcohol is bad for children or not? Not possible
As you can see, the first example above has caused confusion. If the answer is “possible”, it means that London Underground DOES provide a good service. The second example is worse, it is a combination of double barrelled and double negative. Any answers given there would not make any representation. If the response is “possible”, does the respondent agree or disagree?
How do you feel about the following statements?
London Underground is not providing a good service for its customers
Alcohol is bad for children
Ensure your answer is specific in order to improve accuracy and avoid confusion.
What is your income?
a) £0 – £1000
b) £1001 – £1500
c)1501 – £2000
d) £2001 – £2500
The problem with above is, when you ask someone for income, what do you mean? Do you mean monthly/weekly/daily from salaries/wages/rental? Do you mean just any income including income given by relatives or do you mean all forms of income?
What is your monthly salary?
a) £0 – £1000
b) £1001 – £1500
c)1501 – £2000
d) £2001 – £2500
Getting ready to distribute and send out the questionnaire.
Test your questionnaire
The person that writes the questionnaire will not be able to identify every single mistake or error. Therefore, you should always ask a few people to complete it and ask them to give you some feedback. This is known as a pilot questionnaire. You should encourage them to tell you anything that is vague or not clear about your questions Back to Top
Where to send
Social networking sites – Facebook, Weibo, RenRen etc Emails – Friends and families Survey – Go to a location yourself and ask random people. Make sure you are in random locations and not just stand in one location unless necessary. By this, I mean do not just stand in front of tube station and ask people “What form of transport do you take to work?” Back to Top
Now that you have received the responses, you may group the results and analyse them. We have created an Excel template to make it easier for you to create charts and graphs.
Use charts such as pie graphs or bar charts along with your questions and results to present the results. The human mind can absorb results better through the use of picture (graphs) and we can see the difference between the responses more clearly too.