How to perform correspondence analysis using Hamlet II

By Avishek Majumder & Priya Chetty on October 22, 2018

The last article was about performing a non-hierarchical clustering text analysis using Hamlet II. This article explains correspondence analysis using Hamlet II for text, taking a matrix of context unit profiles as its input. It was created at the time of performing joint frequency analysis.

Correspondence analysis is a diagnostic tool which helps to present the content of a single text or the context unit profiles containing the joint frequency. In simple words, this analysis helps in representing the 2-dimensional matrix containing rows and columns into a single plot.

Joint Frequency analysis needs to be run in order to save the matrix of context unit profiles for further performing correspondence analysis. After saving the matrix (in this case as “new matrix”), one can perform the correspondence analysis in many different combinations of items present in the existing word list.

Steps to conduct correspondence analysis using Hamlet II

  1. On Hamlet II, go to Tool Bar > CA > Correspondence Analysis of text/context profiles.
    Figure 1: Step 1 of correspondence analysis using Hamlet II
    Figure 1: Step 1 of correspondence analysis using Hamlet II
  2. A Correspondence Analysis dialogue box will appear. First, select the file containing context unit profiles (“newmatrix.xpr” in this example). Next, specify the number of dimensions require to be reported in the new pop-up box and then click on “Analyse these words or categories”.
    Figure 2: Step 2 of correspondence analysis using Hamlet II
    Figure 2: Step 2 of correspondence analysis using Hamlet II
Figure 3: Graphical representation of correspondence analysis using Hamlet II
Figure 3: Graphical representation of correspondence analysis using Hamlet II

The above configuration will appear in graphical form in a new window. This graph helps to display and analyze the common plot of different categories & context units. It also helps to separate plots by clicking on “Next” and “Previous” buttons at the top. This graph helps to understand which row or context units mostly refers to which of these 10 columns. The closer the points are to each other them more related they are to each other and also their underlined factor. Highlight each of the labels in the graph by clicking on their data points and on “Refresh” option to cancel the highlighting. Also, the “Grid” option helps to switch between the grid and non-grip plots.  Drag this plot anywhere in the window using a mouse to focus on some specific parts of the graph.

Methods of editing the graph

  1. Click on ‘Data’ to view or edit the attributes of correspondence analysis result in order to make changes in the existing results.
  2. To be able to change and view the graphical representation in a different orientation, use the “Left” and “Right” buttons. Similarly, the “In” and “Out” buttons help to zoom in and zoom out the graph.
  3. “Labels” option can be used to make changes in the displayed labels in the plot. This option not only allows to change the style and size of the font but also helps to restrict the number of characters in the labels and present it in a clean way.
  4. These tools provide basic drawing functions which allow to draw, highlight or write on the displayed plots. “Draw” function allows to write or draw freely on the current plot. “Line” function helps to create straight lines from one point to another using mouse. “Ellipse” is a great tool to highlight the focused labels by simply drawing an ellipse with the mouse after clicking on the function button. Clicking the “Text” option opens a box which allows typing the text to display and change the style and size of the text. Once the text is ready, hit ‘Enter’ and click anywhere on the plot to display the text. “Erase” can be helpful to erase any unintentional or wrong parts of the display accordingly. “Line Drawing” option in the toolbar helps to change the colour of the lines.
  5. Click on ‘Save’ after editing a graph.

MDPREF using Hamlet II for singular value decomposition

This article focuses on performing correspondence analysis using Hamlet II in order to create and represent 2 or more dimensional matrices in a single plot. It also helps in visually distinguishing different words or contexts on the basis of their belongingness to the relative keyword or column or factor. The next article uses the same matrix of profiles or context units to perform Singular Value Decomposition (MDPREF) in order to represent the number of subjects of preference of each group of stimuli.

Discuss