Steps to conduct MDPREF using Hamlet II for Singular Value Decomposition (SVD)

By Avishek Majumder & Priya Chetty on October 24, 2018

In the previous article, Correspondence Text Analysis was performed to represent 2 or more dimensional matrices in a single plot according to the context’s association with their relative factor. This article talks about the application of Singular Value Decomposition (SVD) technique MDPREF using Hamlet II. It is performed on the same matrix of profiles or context units saved while performing joint frequency analysis. Similar to the correspondence text analysis, SVD (MDPREF) helps to represent the number of subjects of preference of each group of stimuli.

Save a matrix of context unit profiles from joint frequency analysis before moving forward to SVD (MDPREF). After saving that (in this case as “newmatrix”), one can perform SVD (MDPREF) in many different combinations of items present in the existing word list without starting the process all over again.

Steps to conduct MDPREF using Hamlet II

Follow the below steps to perform MDPREF using Hamlet II.

  1. Go to Tool Bar > SVD > Correspondence Analysis of text or context profiles as shown in step 1 of the figure below.
  2. Click on the browse button (….) and select the file containing context unit profiles as shown in step 2 of the figure below.
  3. A new box will appear, which will show the words/categories/groups/factors available in that particular matrix. Select the number of dimensions you are willing to present your data points on the graph or plot as shown in step 3 of the figure below).
  4. Click on “Include these word tokens or topics” button, in order to see the graphical representation as shown in step 4 of the figure below.
Figure 1: Steps 1, 2, 3 & 4 for performing MDPREF using Hamlet II
Figure 1: Steps 1, 2, 3 & 4 for performing MDPREF using Hamlet II

Graphical presentation of MDPREF

Figure 2: Dimensional graphical representation
Figure 2: Dimensional graphical representation

Dimensional graphical representation of the data points and their underlying factors appears after running SVD text analysis. Similar to the correspondence text analysis, the context units are plotted based on the relativity weights or the degree of association they carry in accordance to all these tokens or factors. For an instance, the above graph shows that U114, U135, U206, U139, U82 etc. are closely related to the token “WORD”. Likewise, “TEXT” is mostly associated with the data points U116, U129, U49, U274 etc. One can also drag the above graph anywhere within the window to closely analyze each data point.

Exploring the data points from the plot

Figure 3: Exploring the data points
Figure 3: Exploring the data points

To explore and analyze the data points in this 2-dimensional graph, click on the ‘Data’ option on the toolbar and click on ‘View or Edit data’. Data points marked in red are closely situated with WORD and the ones in blue are defining TEXT. This further confirms existence and the degree of relationship in the graphical representation of these data points.

Most of the options available above are same as those in the correspondence text analysis article.

  1. Options like “Left” and “Right are available in order to change and view the displayed plot in different orientations. Again the “In” and “Out” functions help zoom in and zoom out to closely examine each of the data points on this graph.
  2. Clicking on a specified data point will highlight the particular label by enlarging and changing the colour of it. To reverse the action, use the “Refresh” button from the above toolbar.
  3. The other option “Labels” is also available so as to change and view the labels in different styles and sizes. It also helps in shorten the number of characters used in labels to make it look less clustered.
  4. Another useful tool here is “Vectors”, which helps to toggle between the plotting of vector lines. Turning it off can make the graph look cleaner.
  5. Editing in Singular Value Decomposition.
  6. “Grid” option helps to toggle the displayed grids on the graphical representation
  7. If satisfied with the displayed graph/plot, click on the Save Display option to save the edited plots

Steps for editing in Singular Value Decomposition (SVD)

Just like the correspondence text analysis, SVD also offers some of the basic drawing tools which can be used to draw, highlight or write on the displayed plots. “Draw” function allows to draw freely with the help of mouse on the displayed plots.

“Line” function helps to draw a straight line anywhere on the plot. “Ellipse” tool helps to create an ellipse shape using a mouse to highlight a particular group of data points on the graph. Clicking on “Text” button, a new pop-up will appear, allowing you to place text anywhere on the plot. “Erase” helps in erasing any part of the currently displayed graph. The Line Drawing option in the toolbar above allows making changes in the colour of lines drawn.

Non-metric and metric multidimensional scaling analysis using Hamlet II

This article revolved around the application of SVD (MDPREF) using Hamlet II 3.0. This analysis helps in the graphical representation of the data points according to the association of each underlying factors or tokens. This analysis plots a clean graph and also shows the related matrix to clarify the strength of the relationship between two data points. The next articles focus on applying non-metric and metric multidimensional scaling methods using Hamlet II. They are useful in studying the joint occurrences of words, and in comparison of words from different transcripts.

Discuss