Method

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 
136 137 138 139 140 141 142 143 144 145 146 

Using the four-stage process detailed above, we conducted two interlinked studies

(Armstrong 2001): Study I, incorporating stages 1 and 2 of the validation process (see

Figure (1b)), used interview data gathered from expert developers to elicit Revealed

Casual Maps (RCMs). Study II gathered survey data from a large sample of software

developers to unearth the empirical structure of the raw concepts from interviews and

to reconstruct the RCMs through bootstrapping, stages 3 and 4 of the validation process

(see Figure 1(b)).

Study I: Revealed Causal Mapping

To represent the knowledge structures of OO software development experts, we followed

the causal mapping procedure detailed in Chapter II.

• Data source. For this study, the data source was domain experts and the narratives

were gathered through semi-structured, open-ended interviews. To accomplish

this task, expert OO software developers were identified using a snowball technique

(Shanteau, 1987, 1992) and convenience sample (Stone, 1978).

• Sample. The participants in the study were expert OO software developers, as

acknowledged by their peers. Organizations were selected based on their identification

of available “expert software developers” and their willingness to participate.

Over 15 organizations of various sizes (15-10,000 employees) and industries

(e.g., telecommunications, manufacturing, consulting, and services) provided

access to their software developers. In Table 3, we summarize the key characteristics

of the respondents.

• Interviews. The interview process consisted of open-ended interviews with probes

(Rossi, Wright & Anderson, 1983). An interview guide was adapted from Nelson,

Armstrong and Ghods (2002) to facilitate the interview process (see the Appendix

for the Interview Guide). During the interviews respondents were asked questions

regarding how they think about software development. Based on the respondent’s

answer to the question, follow-up probes were asked to elicit further details

regarding their software development thought process. Each interview lasted from

30 to 90 minutes. The range of interview lengths occurred because the interviewer

did not constrain the responses to the questions. The interviews were tape

recorded and later transcribed into a document format ranging from four to 14

pages.

Prior to commencement, we estimated the number of interviews necessary to reach

redundancy or saturation of concepts at 25. The point of redundancy was reached at 15

participants eliciting a total of 19 concepts. The point of redundancy suggested that the

achieved sample of 24 respondents was more than sufficient to capture all of the relevant

concepts in the sample.