4.3 Development of a taxonomy
4.3.2 Planning the Iterative classification process
This iterative section of the development process begins with Step 8. If in Step 12 it is determined that the ending conditions have not been met the process will return to Step 8, otherwise the classification process ends.
Step 8: Determine classification approach (Nickerson et al. 2013 & Usman et al. 2017)
The nature of each of the two taxonomies identified as a starting point are considered here, and how that impacts on the classification process. Finally, the approach used for the first iteration is discussed.
Implementation Taxonomy
It is intended that initially an inductive process is employed (Nickerson et al. 2013, p.334), being derived from the empirical and generalised to the conceptual across all the studies to realise column headings for a matrix. Each paper will be read and the
characteristics for classification into a facet will be noted in the matrix.
attempted to trial different things within the same reported study. A decision had to be made regarding how to classify the data whilst maintaining clarity. One option is to provide clarification by separating out the studies into two or more studies, which would require an additional identification characteristic for the separate studies, in such a way that also identified them as related. Another option is to treat the publication as one study and to enter multiple values in some facets when required. The latter was dismissed since it was impossible to work out how practice was implemented when there are multiple values for the same facet. For example, when you look at Table 4.3 is it possible to answer the following questions?
• Which assignment type was screen cast and which was recorded with the camera?
• How many students completed assignments of each type?
Therefore, when the information diverges at a single facet within the same study it will be split into two or more distinct entries in the taxonomy, even if that means much of the data is repeated. This will require a distinguishing facet being used to identify the different groups of results from the same study.
Subject Recording Source Assignment type No. of students
Industrial Revolution
screen cast camera
essay
presentation 232
Table 4.3: Example of how entering multiple values for a single facet, as characteristics of the same study, would reduce clarity
Perceptions Taxonomy
The facets in this taxonomy are conceptual labels for comments which make pertinent points regarding perceptions. Familiarity of the text means it is already known that sections of the text may contain overlapping points in the same paragraph or few sentences. Chopping phrases into increasingly small sections to prevent overlapping sometimes results in the loss of important contextual information, and sometimes division may be deemed impossible. Therefore, it was decided that comments may need to be repeated in different facets of the taxonomy if they were used as evidence of different points. An alternative approach is to classify general comments at a more general level.
Approach
Some iterations will follow the inductive approach, also chosen as a starting point, while other iterations will apply a deductive approach, moving from the conceptual to the empirical. This aligns with the recommendations of Nickerson et al. (2013, p.345) who suggests that different approaches are used with different iterations to ensure no new insights are missed.
Step 9: Define units of classification (Usman et al. 2017)
Usman et al. (2017) suggests defining ‘descriptive bases’, or a set of attributes used to classify subject matter instances. These are referred to by Nickerson et al. (2013) as ‘dimensions’, which in this faceted taxonomy translate to ‘facets’.
According to Nickerson et al. (2013), the classification process is broken down into two separate branches which differentiate between the two approaches selected for the current iteration (inductive or deductive). Each branch consists of three steps. One step is the identification of new items (whether characteristics or dimensions, depending on approach), another regards grouping, and finally there is the revision of the taxonomy. It was quickly realised that if this process was followed to the letter the taxonomy always grows, as facets are never merged or removed. Although these options are briefly mentioned, the possibility of taking steps which may reduce the taxonomy size is not included in the diagram. Therefore, those steps are not explicitly followed here. The substeps considered useful here are :
• Optionally, classify characteristics to facets (ie: if approach is inductive).
• If a required change is identified to the facets, in terms of, creation, deletion, merge, rename or reposition in the structure, the taxonomy should be reorganised.
• Ensure no pertinent information was lost during any changes by reclassifying all data from moved or eliminated facets.
• If new facets were added during the reorganisation, previously examined objects may need re-examination to determine if characteristics exist for classification under the new facet.
Whilst taking care not to allow the taxonomy structure to grow too large, it is important to balance this requirement with ensuring a comprehensive taxonomy. The taxonomy of video feedback should be ‘collectively exhaustive’, (Nickerson et al. 2013, p.346)
meaning that everything deemed pertinent should be classified. Therefore, additional facets may be created to sufficiently explain characteristics of practice, providing enough information to be useful to practitioners and researchers. Reconceptualisation of facets may mean that some are combined or deleted.
Step 10: Revise taxonomy (Nickerson et al. 2013, p.343)
Nickerson et al. (2013, p.343) positioned this step at the end of each of the two
branches for both inductive and deductive approaches. Therefore, it occurs when either approach is implemented. A new version of the taxonomy is developed and may be re-diagrammed at this point.
Step 11: Revisit meta characteristic (Nickerson et al. 2013, p.343)
Since the meta characteristic may not become clear until later in the classification process, this explicit step will ensure the meta characteristic is reconsidered at this point.
Step 12: Determine if ending conditions have been met (Nickerson et al. 2013)
The ending conditions selected in step 6 will be examined at his point to determine if they have been met. If they are all met development will cease, and validation will begin. If they have not all been met the classification process will be repeated from Step 8 until the ending conditions are met.