Testing of a Theory


Fig. 9.1

The process of developing a hypothetical model for theory testing




Table 9.1

Evaluation criteria for the statistical testing of the theory (Acton et al. [11])






































Conceptual framework of the theory


 1. The theory is described precisely, including structure and concepts


 2. The research questions and hypotheses are logically derived from theory and evidence


 3. The research questions and hypotheses are sufficiently detailed to enable assessments of theory validity


 4. The operational definitions of concepts are clearly derived from theory


The study (data collection and analysis)


 1. The research design is appropriate for the type of theory


 2. The instrument/s is/are reliable and based on theory


 3. The theory guides participant selection


 4. The statistical methods used when testing the theory correspond with the type of theory


 5. The statistical analyses provide empirical evidence that supports, refutes, or modifies the theory


Results


 1. The research report includes an analysis of the empirical results related to the tested theory


 2. The research report discusses the importance of the theory for nursing


 3. The theoretical conclusions are used to make recommendations for future research


 4. Theory testing is mentioned in the title, summary and keywords of the research report



9.3 The Process of Theory Testing


9.3.1 Aim


The purpose of theory testing is to verify the validity of a presented hypothesis about the theoretical structure of theory in empirical reality [3, 10]. A theory should be corrected or completely rejected if it does not receive support from the empirical data. A theory can be considered valid when the presented hypotheses gain empirical support. A theory should be tested on a continuous basis, and preferably with different target groups. A theory has a higher degree of validity as the hypotheses amass more empirical support [12, 13].


9.3.2 Study Design


The research setting used to test a theory depends on the type of theory that is being tested. Descriptive theories are tested using a descriptive study design; explanatory theories are tested using a correlative study design; predictive theories are testing using an experimental study design; and guiding theories are tested using repeated measurements and interventions [7, 12]. Descriptive and correlative study designs define the relationships between the concepts described in theory, but they cannot be used to identify the causal relationships between concepts. For example, correlation coefficient between two variables does not describe the structure of the theoretical model, and thus, cannot be used for further theory development [12].


Explorative study designs examine the relationships between the concepts identified for a certain phenomenon, for example, by suggesting causal relationships. The experimental study design allows accurate descriptions (direct or indirect causal relationships) of the relationships between concepts [7] and shows how a change in one factor affects the factors included in the tested theory. It is important to determine which analytical methods will provide the best evidence of the verifying of a theory. A researcher can start the testing process with a descriptive study design and, once evidence for validity has been acquired, extend the testing to an experimental study design [3, 7, 9, 12].


9.3.3 Data Collection


Data will be collected through either direct or indirect observations, such as surveys, interviews, observations and objective measurements. The target population should be representative of the group or context to which the theory is applicable. The sample size can be calculated by power analysis according what has been presented in previous studies that were conducted in the same or sufficiently similar context [12, 1416]. Researchers will often develop an instrument that measures the concept(s) presented in the theory before the statistical testing of theory. The instrument will have to be pretested and psychometrically tested before the hypotheses are empirically examined [16]. Chapter 8 provides more detailed information about instrument development.


9.4 Data Analysis


Statistical methods are commonly used to test explanatory, predictive and guiding theories to draw conclusions about the hypotheses being studied [12]. In particular, factor analysis and structural equation modeling (SEM) have used for testing theories [13, 14, 16]. SEM combines both factor and regression analyses. It allows the study of causal relationships between factors by using regression analysis [17, 18].


Explorative factor analysis (EFA) and confirmatory factor analysis (CFA) belong to the ‘family of factor analyses’. EFA is used to determine exploratory factor model without an a priori assumption of associations between variables. No hypothesis of the factor structure of the data is needed to use it. Based on EFA researchers know how many factors the variables are intended to form which variables are loaded on which factors and whether the factors are interrelated. After EFA, CFA can be conducted to test nursing theory that has already been established. Researchers have to have an a priori hypothesis based on theoretical knowledge or empirical indications [18].


The theoretical basis of CFA relates to fundamentals of SEM. It describes the relationships between variables. The phases of CFA can be represented as preparation and model testing [18]. The preparation phase, which precedes the testing of a theoretical model, is concerned with the quality of the data. During this phase, the researcher will test their data for missing values, univariate and multivariate outliers and normality (for a description of data quality, see Chap. 8, testing an instrument’s psychometric properties). Furthermore, instrument validity should be confirmed with exploratory factor analysis, and more preferably, with confirmatory factor analysis (CFA). Various statistical cut-off values for goodness of fit can be used to evaluate whether a tested model is valid. Some of the most commonly used cut-off values for goodness of fit include: Root Mean Square Error of Approximation (RMSEA) <0.08; Standardised Root Mean Residual (SRMR) <0.08; Comparative Fit Index (CFI) >0.90; and Tucker-Lewis Index (TLI) >0.90. The CFA of an instrument will yield observed variables (which describe the items of an instrument) and latent variables (which describe factors that explain a minimum of three observed variables). An example of a CFA result for instrument validation is shown in Fig. 9.2.

../images/470474_1_En_9_Chapter/470474_1_En_9_Fig2_HTML.png

Only gold members can continue reading. Log In or Register to continue

Apr 18, 2020 | Posted by in NURSING | Comments Off on Testing of a Theory
Premium Wordpress Themes by UFO Themes