NEDV with Pattern-matching for RH1 Hypothesis Testing

This study applied the NEDV with pattern matching (T. D. Cook et al., 1979; Trochim, 1989).

Trochim comments that the NEDV alone can be rather weak with respect to internal validity.

But in combination with pattern matching, it strengthens the design for causal assessment and becomes extremely powerful. The NEDV can have a single dependent variable and thus a single group of participants. This single dependent variable design can have more than one dependent variable. Thus two or more dependent variable design would be made considerably stronger because of the higher level of internal validity and the multiple outcomes from several dependent variables. This is briefly described in the following.

The NEDV with pattern -matching was described by T. D. Cook et al. (1979). Trochim (1989) describes that the theory of pattern matching will help researchers to develop theory to a greater height. He states that pattern matching emphasizes the importance of theory driven approach in applied research. This importance has been reemphasized by concerted efforts (Campbell, Stanley, & Gage, 1963; H. T. Chen & Rossi, 1987). Trochim commends favourably Chen and Ross for critical view on a research paradigm which has encouraged an experimental perspective at the expense of richer theoretical elucidation. Pattern matching has a long tradition of research thinking, and was originated from Chen and Rossi in 1987.

Among the concerted efforts, Ferrari et al. (2008) address the construct validity of measures

104 by developing a “nomological network” to link theoretical propositions and constructs with operational definition as part of the inferential process.

The theory of pattern matching (Trochim & Cook, 1992)states that pattern matching specifies a predicted pattern from a theoretical framework, acquires an observed pattern from an empirical study, and attempts to match these two patterns. If the two patterns match, the results can help to make stronger the internal validity of the case study for the quality of research. The general idea of pattern matching is illustrated in Figure 3.18.

Theoretical Realm

Observation Realm

Figure 3.18 Pattern Matching Logic

There are two activity realms comprising theoretical realm at the top and observational realm at the bottom. In the theoretical realm at the top, there originates the theory which might come

Theoretical (Predicted) Pattern Conceptualization Task

Observed Pattern

Data Organisation - Theories - Ideas - Hunches

- Observations - Data

- Measures

Pattern Matching

105 from a formal tradition of theorizing or the ideas by researchers. These ideas are then translated and specified into a theoretical pattern as shown in the oval shape through the process of conceptualization. Trochim states that a theoretical pattern is a hypothesis about what is expected in the data, while the observed pattern is the acquired data which are used to examine the theoretical model.

In the observational realm at the bottom, there will be the direct observation of study objects or entities at real or controlled environment in the data form of field or lab notes, and formal objective measures. The collected data will be analysed and organized through an operational definition process, becoming the observational pattern. The definition is vital for the relevant linkage of the observational pattern to the theoretical pattern.

At centre of the figure is where the inferential role plays to match the two patterns. If the patterns match, a conclusion can be drawn to support the theory which might predicts the same observed patterns. If the patterns do not match, a conclusion can be drawn that the theory may be not correct or the observation may not be correct. The evidence provides for internal validity of the case study research (Herriott & Firestone, 1983).

The NEDV with pattern matching was applied to test hypothesis RH1 that architectural design quality decreases as market price decreases as shown in Figure 3.21. This test presented analytical generalisation. With replication logic and inferential statistical testing as described in next sections 3.6.4 and 3.6.5, statistical generalisation for the role of theory development is obtained as shown in Figure 3.19.

106 Predicted Patterns Observed Patterns

N O1 X1 O1

O2 O2

O3 O3

O4 O4

O5 O5

O6 O6

O7 O7

Figure 3.19 NEDV with Pattern Matching for testing RH1

• NEDV with pattern-matching logic provides a test for the hypothesis RH1. If the predicted patterns and observed patterns match, the results can strengthen its validity.

The predicted patterns are the hypothetical proposition while the observed patterns are the empirical evidences of the subjects under observation study.

• There are seven dependable variables (DV), hence a variety of seven expected outcomes that are represented by O1 to O7. The seven DVs are the seven design criteria. The first DV1 = site context criterion given a market price, and DV2-7 are the other six design criteria.

• Analytical generalization is used for an inference or a priori prediction from cases on the basis of empirical data collected, while statistical generalization is used for the said purpose from survey samples on the basis of empirical data collected.

• The results are stronger if there were multiple outcome variables as proposed in this study or a priori prediction of extent to which each will be effect.

• Previously developed theory or the predicted theoretical pattern is used as a template with which to compare the empirical results of the case study.

• If two or more cases support the same theoretical pattern, replication may be claimed that can strengthen its validity

• Results may be considered more potent if two or more cases support the same theory or predicted theoretical pattern.

107 3.6.3 Replication Logic

The logic of replication (Yin, 2011) is applied for this study. The number of case replications is the essence of this logic rather than size of case sampling. The logic has five steps as shown in Figure 3.20: firstly, the theory development; secondly the selection of cases and the setting up of specific measurement tools in the data collection; thirdly the conduct of every individual case as a ‘complete’ study and the conclusions for each case as the required information for replication by other individual cases; fourthly the analysis of the data; fifthly the drawing up of cross-case conclusions on the proposed theory etc.

DEFINE & DESIGN PREPARE, COLLECT & ANALYSE ANALYSE

CONCLUDE

Figure 3.20 Replication Approach of the Case Study Method (COSMOS) Select

108 Hersen, Barlow, and Kazdin (1976) comment that replication logic is similar to that in experiments. In a single experiment, a researcher would discover a significant finding of the first experiment that will replicate by repeating a second, third and more experiments. To validate the results, the researcher would conduct some replications using the exact conditions of original experiment, while other replications by changing conditions to the original experiment.

The replication of the case study in multiple-case studies follows the similar logic of experiment. A case study researcher will select and conduct about seven cases. Each case is liked, conducting each experiment, will either predicts similar results or opposite results. The predicted similar results are a literal replication while the opposite results a theoretical replication. If all cases were achieved as predicted, these cases would give significant support for the study propositions. Failing which, the initial study propositions must be corrected and tested again with another set of cases.