Analysis method of the data was most difficult to decide, since I was not very familiar with quantitative research methods such as statistical analysis before.
Therefore, I had to do a lot of reading until I was able to decide my statistical analysis technique and overall to understand how statistical analysis is done. I was familiar with excel and eventually started my analysis with it, since it felt easy at first to approach my data with it.
At first, I transferred my survey data on Excel by counting how many answers there were on Likert’s scale for each question to see the percentages of each question’s answers. I also formed tables out of so-called personal questions that were analyzed with excel as well. These informative questions can be called as subjective descriptors because these categories are chosen because of their probable relevance to topics of the survey. (Gillham 2000, 49.) First, I formed analysis grid where you can see all the subjective descriptors of each answer sheet. This helps me to find personal data of each participants quickly. Secondly, I formed grids of every personal detail, and produced chart of each category in order to see overall percentages of each subjective descriptors.
However, I soon noticed that I won’t be able to get much analyze out of this kind of graphical display, so I started to find other ways to analyze surveys that are mostly concentrated on finding out people’s opinions and are measured by Likert’s scale.
Scholar Bill Gillham’s Developing a Questionnaire (2008) turned out to be really helpful guide for the analysis of questionnaire research method. The book was firstly published in 2000, and both of the printouts are used as guidebooks. According to Gillham’s (2008, 50) advise, I prepared numerical table (data matrix) out of my answers on excel. I conducted more graphical displays of out of my data as well as cross tabulating on SPSS. However, I wanted to perform statistical test on my data as well. Non-mathematicians can only have limited use and understanding of statistics, but even with this limited use, non-mathematicians can gain useful knowledge with certain simple tools and make better sense out of their data – to say more and to say it validly. Gillham (2008) suggests the usage of chi-square, because it deals with categorical data, data that is categorized in certain themes or groups. Chi square also enables researcher to say something more valid about differences between different
33
categories such as gender. (Ibid., 71.) However, after reading lots of articles and comments about which statistical test I should perform for my data, I decided to use Mann-Whitney U and Kruskal-Wallis tests instead.
The usage of Mann-Whitney U test and Kruskal-Wallis test was decided, because the data in non-parametric18, variables are ordinal19 and I want to know if there is difference between different categories in how they answer to certain questions.
Mann-Whitney test is used to test, whether there is significance difference in answers between two random samples (like gender). Since my survey is researching attitudes and is measured by Likert’s scale, I cannot perform t-test. Kruskal-Wallis test is similar to Mann-Whitney U test, but in Kruskal-Wallis there can be more than two samples (university degree, ideology). (Tikkanen 2017, 20, 25.) Both of test are based on null-hypothesis, which means that researcher will not find any difference between two groups or in other words, answers are similar among all groups. P-value (probability value) that is used is 0.05 (5 %), which means that if the given results is below 0.05, probability is significant or in other words, researcher can reject the null hypothesis. That is, there is difference between the samples. (Gillham 2000, 87, 100.) The test is run with SPSS, which I was able to download on my computer and I transferred excel data matrix on SPSS. I had not used SPSS before, but with help of and small SPSS guide from University of Turku by Tikkanen (2017) I was able to conduct my analysis easily. I got access to this guide with help of my brother, who served as my statistical analysis mentor during the research process. Furthermore, Surveying a Social World – Principles and Practice in Survey Research by Aldrige
& Levine (2001) serves as helpful guide book as well, since it explains the survey research process very precisely from the beginning to analysis and presentation of the results.
Open-ended questions are analyzed differently. First of all, the questions had to be translated into English. This caused a lot of trouble, since sometimes it was hard to understand what participants had written because of different handwritings, especially since I did not have much experience of reading hand written Korean.
18 Non-parametric statistics rest on relatively few assumptions about the population where the sample is drawn (Aldridge, Alan & Levine, Ken 2001, 179).
19 Variables that are on ordinal level of measurement form a ranking or in other words, categories have intrinsic order (Aldridge & Levine 2001, 130).
34
Once again, I used Google translate as my main tool. The answers of which I was not completely sure what the participant meant I decided to ignore in order to avoid misunderstandings. Once I had translated the answers, I wrote them down to my computer in English and started analyzing. As mentioned before, all four questions are divided into four themes. However, according to Gillham (2000, 66-67), answers should be divided to different categories as well and that when going through each answer, it is necessary to highlight substantive statements, those that really say something. Therefore, as for method, thematic analysis is used for open questions.
An article by Moira Maguire and Brid Delahunt (2017) is used as methodological guide. Thematic analysis identifies patterns or themes within qualitative data.
Attention is paid for themes that are interesting and say something about the research issue. However, good thematic analysis looks beyond what people have said.
(Maguire & Delahunt 2017, 3352-53.) In the next chapter, results of the analysis are presented.
35