CHAPTER 4. METHODS 21
2. Data Collection for the Quantitative Survey 31
Pre-Test for Survey
Before the surveys were distributed by the Directors of TECO, the researcher went to the Culture Center of TECO office in Boston and asked for seven Taiwanese immigrants who met study criteria who would be able to fill out the pre-test survey. The first pre-test was on June 2, 2013. Comments from these respondents helped me to revise the survey questions. The final survey questionnaire was written in English and includes a translation in Traditional Chinese (Mandarin).
Survey Distribution and Completion
The Institutional Review Board (IRB) at Boston University approved this research proposal on July 25, 2013. The online survey was conducted from July 26, 2013 to September 8, 2013. The mail survey and distributed survey occurred between July 27, 2013 and September 20, 2013.
The survey was conducted primarily by mail, online and distributed through agencies. In order to get enough and diverse Taiwanese-immigrant respondents who met study criteria in the East, Central and West, I contacted the Culture Center of Taipei Economic and Cultural Offices (TECO) in the United States in Boston, New York, Chicago, Houston, Los Angeles and Seattle. The directors of the Cultural Center of TECO in the above six areas are familiar with and in close contact with the major Taiwanese associations in each area; for example, some local affiliates and associations for Taiwanese immigrants, churches, religion association and senior groups. I had
previously contacted all of the TECO directors and explained the purpose of the study and the study inclusion criteria, and the directors agreed to help me complete the study. Each director assured me that there are ample numbers of people in each area who met my study criteria.
Directors offered to help in a variety of ways: 1) to mail the survey to Taiwanese immigrants directly who meet study criteria, or to enlist the help of local Taiwanese associations and agencies in their areas to mail the survey to Taiwanese immigrants; 2) to email the link to the online survey (I provided the link to the survey) to prospective respondents who met the study criteria; and 3) to distribute printed copies of the
questionnaire to local Taiwanese associations and agencies which would distribute them to Taiwanese immigrants who meet study inclusion criteria.
To be included in the analysis, respondents had to meet the following criteria: 1) The respondents must have agreed to participant in the survey by online or
mail informed consent.
2) The respondents must self-report that they come from Taiwan before completing the survey.
The total number of surveys obtained for the quantitative analysis was 605. This included 64 completed mail surveys, 152 who completed the online survey, and 389 who completed the paper survey distributed by the directors of TECO and myself. Details on the collection of survey responses through each mode of administration are described below.
Mail Surveys
I distributed 200 questionnaires by mail in the areas represented by the Culture Center of Taipei Economic and Cultural Office (TECO) in Boston, New York, Chicago, Houston, and Los Angeles. The number of completed mail surveys totaled 64. The response rate for this mode of administration is 64/200, or 32%. Directors at each TECO mailed copies of the survey directly to Taiwanese immigrants in their areas or distributed the survey to area Taiwanese associations to mail copies of the survey to prospective participants who meet study inclusion criteria. Each survey that was mailed to a potential participant included the recruitment information, informed consent, and a self-addressed stamped envelope (SASE) so that the survey could be returned easily.
Online Surveys
The online survey questionnaire in English and Traditional Chinese was placed on the Boston University Qualtrics website. I distributed the link to access the survey to the directors of the Culture Centers of TECO. The directors directly sent the link to
Taiwanese immigrants or sent the link to the local directors of Taiwanese-American associations in Boston, New York, Chicago, Houston, Los Angeles, and Seattle, which in turn emailed the survey link to prospective participants. Overall, 224 online surveys were distributed. Each potential participant in the online surveys was contacted up to 3 times, a one-time mailing of the survey and two reminder emails were sent to online participants with the link to the survey. Among 224 who started the online survey, 152 agree to participant in the survey and completed online survey within the period of survey,
so the online response rate is 152/224=67.8%. The email with the survey link included recruitment information and the online survey included a copy of the informed consent.
Distributed Surveys
In Boston, New York, Chicago, and Los Angeles, I arranged with the local TECO directors to personally distribute the survey questionnaire at the Culture Center of TECO and to members of local Taiwanese associations, churches and non-profit agencies. In some cases, respondents completed the survey using printed copies that I distributed when visiting a Taiwanese immigrants’ association. In other cases, the agencies had Taiwanese immigrants complete the survey. In these cases, I worked with local agencies to be sure they understood the study inclusion criteria and agreed to send completed questionnaires back to me. The distributed surveys also clearly stated the study inclusion criteria. The agencies collected the completed questionnaires and mailed them back to me. In all cases, the printed questionnaire included the recruitment information,
informed consent, and a return address so that the survey could be returned easily to me by the association or agency. Overall, I distributed 850 surveys in this way. A total of 389 printed surveys were completed, resulting in a response rate of 389/850=45.7%.
Missing Data
Of the 605 surveys obtained, 22 cases were deleted because of missing substantial data. In these cases, respondents answered less than 50% of questions on the survey or were missing responses to 17 or more questions in the survey. As a result, the total valid sample size for the quantitative study was 583 cases.
The Survey Instrument
1) Questionnaire design—some questions were difficult for some respondents to answer. For example, several questions asked respondents to rate the
importance of an item the question “I have a new job in Taiwan”. However, a review of surveys where these data were missing suggests that respondents who did not have jobs in Taiwan did not think it was necessary to address the importance ratings. Another question, “My health condition would not allow me to take a long flight” also resulted in cases where respondents did not answer the importance ratings.
2) Over 10% of the cases with missing values were due to two questions: “I am retired from a Job in Taiwan” (missing n= 65, 11.1%) and “What is your current annual income (in USD)?” (missing n=60, 10.3%). It is possible that respondents did not know how to answer these questions or were reluctant to answer them. In my qualitative interviews, I learned that respondents were reluctant to answer personal questions about their work status and income. 3) Surveys printed on both sides – some respondents did not complete questions
on the reverse page.
4) Some respondents became tired or did not have the patience to complete the survey.
Analyzing Missing Data
Within the 583 cases included in the quantitative study, a detailed analysis of the missing data was conducted. For each variable that contained missing data, a comparison
was made between respondents who were missing data for the variable with respondents with completed data. All bivariate analyses were repeated with missing cases and
compared with the bivariate results based on completed cases. These data (not shown) revealed no significant differences between respondents who were missing data and those who had completed surveys. To analyze the effect of missing data on the final logistics regression, mean imputation and multiple imputation methods were used. Missing data in category variables were recoded as a new category or combined into the “other” item (Allison P., 2002; Graham, J.W., 2012). Results for these analyses showed that the analysis with imputed data were highly comparable to the analysis based on the completed cases with similar significance on the same variables. For this reason, the bivariate and multivariate presented in this dissertation are based on the 583 completed cases.