抽樣方法案例

4.2. Data collection Our sample was obtained through a simple random sampling

technique. First, a sample frame of 170,000 subscribers with an email address was

provided by the Technical Department of the property developer. Second, each subscriber was assigned a number, ranging from 1 to

170,000. Then, 10,000 subscribers were chosen by matching the number with output from a random number generator.

The questionnaire was posted on the company’s portal. An invitation letter with a hyperlink to the Web page with the

questionnaire was sent to each selected subscriber. The letter explained the purpose of the study and encouraged the

subscribers to participate in the survey to help improve the portal’s service quality.

4.2. Data collection A total of 2120 subscribers responded within a week. After

data screening, we eliminated 128 incomplete and repeat questionnaires. As a result, the total effective sample was 1992 or 19.92% of the sampled subscribers. As all the questionnaires were collected within a week, we did compare earlier and later responses. The non-respondent question was examined by comparing respondents’ demographic variables, i.e., age, sex, marital status, and education levels, with those of non-respondents, which was available from the portal’s subscriber database. A t-test indicated that there were no significant differences. Thus, our data were suitable for further analysis.

4. Methodology

Companies included in this study were selected from the list of the 500 largest global corporations published annually by Fortune magazine. This source was used because such companies provide leadership in the use of information technology [14], [15], [16] and [17], are leaders in their industry, have the resources to develop and maintain fully-featured web sites, and represent a true global mixture of organizations. 

6. Methodology Purposive sampling was employed. This approach

obtained an overall frame of potential respondents and then created a sub-frame of respondents with desired characteristics.

Such designs are considered entirely appropriate in explanatory studies that examine unique or complex phenomena [53].

The sampling frame adopted was the East Edition of The Directory of Top Computer Executives [1]. This contains the names, titles, addresses, and phone numbers of top computer executives in the Eastern half of the United States. The entities within the directory include Fortune 2000 firms (manufacturing and service), educational institutions, hospitals, and governmental agencies.

6. Methodology In developing the desired sub-frame, all hospitals,

educational institutions, and governmental agencies were eliminated because they have entirely different success factors.

Next, the job titles of respondents remaining in the frame were examined to determine their level of planning activity. Firms whose senior IS manager had the job title of: Chief Information Officer, Vice President, Director of Strategic Planning, or Director of MIS were retained.

The resultant sub-frame consisted of 1100 business entities. From this, 600 firms were chosen at random.

3. Research design

As a sample frame we used two directories: Yahoo (www.yahoo.com) and the Dutch Yellow Pages (www.markt.nl).

The directory ‘Business and Economy – Companies’ of the Yahoo Guide provides one of the best known international lists of links to commercial Web sites; the majority of its companies are from the US.

The Dutch Yellow Pages contains companies from the Netherlands only.

Altogether, we analyzed 651 Web sites, 501 from Yahoo and 150 from the Dutch Yellow Pages.

3. Research design To select companies from both directories, we

applied a sampling procedure that was a combination of quota sampling and proportionate stratified sampling.

The first was used because of the dominance of computer (related) firms on the Internet. A random sample would mostly contain IT-companies. We selected 25 categories (or quota’s) from the Yahoo directory, see Table 1.

If possible we selected a minimum of 15 sites per category (10 for Dutch Yellow Pages). The remaining sites were selected as a proportion of the size of each category.

3. Research design

To select sites within a category, the stepsize s was computed as

where C is the category size, the number of sites in a category, and n is the number of sites from that category that must be included in the sample. Within a category each sth site was selected, until n was reached.