View source for Categorization and pattern spotting (analysis)

'''Categorization and pattern spotting''' is a family of [[Analysis method|analysis methods]] that are used to identify commonalities among cases in data. Both quantitative and qualitative approaches can be used to find commonalities between cases. 

== Definition==
'''Categorization''' is an approach that puts a single [[data]] [[Case (data)|case]] into a category, or identifies the existence of categories from among multiple cases. Cases with similar traits are placed into the same category. 

'''Pattern spotting''' is a [[System / Systematic|systematic]] approach to finding patterns or sequences among data. Cases (or ''people'' when you consider a data case as a single person) that have similar patterns represented in the data may also [[Behavior of the learner (KAB)|act]] similarly, and subsequently have the same outcome (such as [[Learning objectives and outcomes|learning objective achievement]]). The presence of patterns could indicate relationships or similarities between cases, which could be of use to evaluators.

==Additional Information==
In this method, a ''case'' is a single data example that is used for analysis. A case can be:

* '''A single interaction''' in a learning environment with a timestamp (such as watching a video, a dialogue between two students, or submitting an assignment)
* '''A single person''', with multiple variables that represent their observed behaviors, observed performance and achievement, affective states, and background information

The size and scope of the case is determined by the researcher using the method, as well as which other variables or information to include about the case.

As with most analysis methods, the quality of the results from the method will depend upon (1) the quality of the data that are collected from the instruments that are used and (2) the skill and consistency of the researcher in applying the method and procedure. 

'''Common categorization and pattern spotting methods:'''

''[[Quantitative analysis|'''Quantitative methods''']]''

* <u>Logistic regression.</u> This method is used to estimate the probability of a case, given certain variables, falling into a pre-determined category. The categories in logistic regression are already defined before the study begins, but can be a useful tool for determining the probabilities and chance of a new case being in a category given certain other traits.
* <u>Internal reliability analysis.</u> This method examines how strongly related a set of given variables are in that their values are similar between each case. Low reliability results would indicate that the variables do not go well together and they are not correlated. Conversely, high results indicate that the variables are closely aligned and consistently have similar values within each case.
* <u>Factor analysis</u>. This method reduces a large number of variables into a smaller number of variables by finding similarities between variables.
* <u>Cluster analysis</u>. This method examines datasets of individual cases to find patterns of similarities between cases and sort them into "clusters." The statistical analysis does not identify ''why'' the cases are similar - only that they exhibit similar patterns in the values of the variables that were considered. The researcher must interpret the data and identify what the clusters mean.
* <u>Artificial Intelligence (AI) / Machine Learning (ML)</u>. Methods in AI and ML are designed to find similarities and patterns within datasets. Although technical in scope, such approaches can be "trained" to identify and classify differences between cases that are input into the software. 

''[[Qualitative analysis|'''Qualitative methods''']]''

* <u>Multi-phase coding/labeling</u>. These types of studies go through multiple rounds of investigative data [[Coding (analysis)|coding]]/labeling. With this method, researchers generally do an initial round of coding that identifies very specific categories of data. The next round of coding will 
* <u>General thematic analyses</u>. These broad range of approaches [[System / Systematic|systematically]] investigate qualitative data to identify common themes and trends in the data. 
* <u>Word repetitions among cases</u>. This approach looks for patterns of specific word use among cases, which could identify similarity between cases (and thus the presence of categories or themes)
* <u>Constant comparison</u>. In the ''grounded theory'' approach to qualitative research, researchers compare and contrast examples and cases repeatedly (i.e., ''constant comparison'') to identify both commonalities and differences in the data. This could lead to categorization by the researcher as they investigate the data.

Note: It is beyond the scope of this knowledgebase to expand on each of these methods. It is recommended that researchers and evaluators seek additional training, web resources, or courses on individual methods they would like to use.

==Tips and Tricks==

* When answering your [[Research question|research questions]], consider whether it would be helpful to know if there are groups of participants [[Behavior of the learner (KAB)|behave]] differently, or whether there are types of [[Attitudes and affect of the learner (KAB)|affective factors]] that could influence how people interact in the [[learning experience]]. Once you have appropriate [[data]], you can begin to investigate whether different categories of these behaviors or affective qualities exist among your learners to better understand [[Design feature use evaluation|how and why they use]] the [[educational product]], or with what effect these categories have on [[Learning effectiveness evaluation|learning objective achievement]].
* If you would like to use categorization and pattern identification methods, you should seek out additional informational resources, examples, and  courses on these methods. 

==Related Concepts ==

* [[Quantitative analysis]]
* [[Qualitative analysis]]
* [[Analysis method]]
* [[Research question]]

==Examples ==

None yet - check back soon!

==External Resources==

None yet - check back soon!