**2. Methodology**

For this investigation, the analysis reveals a latent underlying structure for the landscape architecture discipline (the landscape architecture research universe) concerning the citation literature of *Landscape Journal* from several decades of articles (1982–2018). *Landscape Journal*, is a preeminent American journal addressing landscape architecture research and is affiliated with the Council of Educators in Landscape Architecture (CELA).

For each issue, the study team collected all of the peer reviewed published articles, 'source articles' for the study years. Each source article comprised one observation set. For one observation set there would usually be journal articles cited in the bibliography. These cited journal articles contained within the bibliography of a source article are called 'citation articles'. To classify a citation article, the Library of Congress classification number for the journal title of each citation article was recorded. If the same journal is cited more than once, the tally will be greater than one. Within an observation set, the total number of citation articles for a particular category was tallied. For example, if the subject category 'architecture' had 6 cited architecture citation articles in a source article, the architecture tally for the observation set would be six. The Library of Congress classification was chosen as it was an existing, broad, and easy to use system, recognized by many major state research universities. The Library of Congress system is non-hierarchical, meaning that the new bodies of knowledge that emerge are relatively easily incorporated into the classification system and thus as the system grows over time, it can accommodate modifications and development in the knowledge base. Flexibility over time was an essential component since historical research may span across a wide time frame.

In this study there were 38 subject variables. Thus each observation sets had 38 scores, each representing the tally of each subject from the source article. With the subject areas for all of the journals identified, one could then sort the citation articles from each source article into a subject category. Citations to literature such as monographs, technical reports, and books were not included in this study. I addition, proceedings were included only if they appeared to be published at least annually, meaning it was a serial. Once the subject areas for each source article were tabulated, the dataset could then be entered into a computer for statistical analysis.

Multivariate data analysis was performed using SAS 9.1 [35]. To conduct a PCA, the subject categories were each standardized to a mean of 0.0, standard deviation of 1.0. The standardization is important to the analysis [36]. Otherwise, the results will be dominated by categories with large scores. After standardization principal component analysis can be conducted upon the observation sets (an observation set is comprised of the scores in 38 subject category variables for a source article). The analysis produces a numerical table present eigenvalues which represent independent dimensions, from the largest value to the smallest. For interpretation, eigenvalues for standardized data with values over 1.0 were considered significant

#### *The American Landscape Architecture Research Universe and a Higher Education Ordination… DOI: http://dx.doi.org/10.5772/intechopen.99119*

dimensions. The significant dimensions represent bodies of knowledge in the landscape research university. Significant dimensions were selected for further analysis by examining the eigenvector coefficients of each dimension which indicate the level of association that a subject category had with the dimension. In other words, eigenvector coefficients numerically illustrate the correlation between a variable (the subject category) and the dimension. The eigenvector coefficients are arranged in a table, sorted by the eigenvalue and would range in score between −1.0 and 1.0. Values near 0 indicate low correlation with the eigenvalue dimension while values near −1.0 or 1.0 indicate a strong association with the dimension. In this study, eigenvector coefficients with a value greater than or equal to 0.400 or less than −0.400 were considered to be affiliated with a particular dimension. Subject categories with more than one significant eigenvector coefficient meant that the subject was significant across more than one dimension, suggesting a dimension connecting subject category. Subject categories with only one significant eigenvector were considered primary to the associated eigenvalue. Primary categories were employed to label (name of identify) a dimension. Weak associations with the dimensions were considered to be eigenvector coefficients ranging from −0.4 to −0.20 and 0.20 to 0.4. The results of the PCA were plotted creating a structural map (universe) of the dimensions, associated subject categories, and connecting subjects. In other words, this map could graphically describe the latent properties of the data. The map would be a graphical depiction of the research universe in a given time frame. Several time frames were examined: the complete time frame from 1982 to 2018, 1997 to 2007, 1999 to 2009, 2001 to 2011, 2003 to 2013, 2005 50 2015, and 2007 to 2017.

In contrast to the research universe, an ordination was also developed describing the curriculum relationships between fifteen top American universities teaching landscape architecture as identified by 'DesignIntelligence,' preparing students for practicing in the profession of landscape architecture [37]. Each school was an observation set and the subjects taught in the curriculum were the categories in each observation set. The categories were standardized, and PCA invoked. The results of the latent dimension can be plotted to illustrate the relative position of one school to another. The intent is not to show which is better, but rather to identify similarities and differences. The plots can depict an educational univers in a manner similar to other types of plots [29–34].
