- مبلغ: ۸۶,۰۰۰ تومان
- مبلغ: ۹۱,۰۰۰ تومان
Concept mapping is now a commonly-used technique for articulating and evaluating programmatic outcomes. However, research regarding validity of knowledge and outcomes produced with concept mapping is sparse. The current study describes quantitative validity analyses using a concept mapping dataset. We sought to increase the validity of concept mapping evaluation results by running multiple cluster analysis methods and then using several metrics to choose from among solutions. We present four different clustering methods based on analyses using the R statistical software package: partitioning around medoids (PAM), fuzzy analysis (FANNY), agglomerative nesting (AGNES) and divisive analysis (DIANA). We then used the Dunn and Davies-Bouldin indices to assist in choosing a valid cluster solution for a concept mapping outcomes evaluation. We conclude that the validity of the outcomes map is high, based on the analyses described. Finally, we discuss areas for further concept mapping methods research.
In conclusion, this study builds additional evidence for the validity of program outcomes articulated via concept mapping. Conducting analyses in the R statistical software package (R Development Core Team, 2011) offers a means of using several methods of cluster analysis. Results reconfirm Kane and Trochim’s conclusion that hierarchical methods prove most useful for concept mapping (Kane & Trochim, 2007). However, R offers the ability to try multiple methods as part of any concept mapping analysis with relative ease. Also, new validation measures are suggested by this study, including the Dunn and Davies-Bouldin indices (Davies & Bouldin, 1979; Dunn, 1974). These assist in choosing among a set of cluster solutions. Similarity indices are also provided (Giurcaneanu et al., 2003) for the purpose of comparing pairs of cluster solutions. All of these indices are provided in the ‘cluster’ (Maechler, 2011) package in the R statistical software environment. Further research into whether and how concept maps could be meaningfully displayed in more than two dimensions and into the issue of scaling raw data would be welcome. In short, even after more than 25 years of using concept mapping, there remain many options and refinements to be explored.