Alignment Analysis

A plane providing a visual representation of how similar each Respondent is, based on how their Comments are coded. A plotted dot exists for each Respondent. Respondents that are plotted closer together have more in common, whereas Respondents that are plotted further apart have less in common.

 

Clicking a Respondent with your mouse toggle's that Respondent's selected status between on and off. Options for selecting groups of Respondents can be found by right-clicking the Alignment Analysis plane.

 

The Filter Group enables you to specify a subset of the project on which you want to perform the analysis. Choosing a Category for example shows you how the Respondents are aligned within that Category.

 

The list of all Respondents that are currently selected in the Alignment Analysis plane.

 

When a Respondent is selected, the corresponding dot on the Alignment Analysis plane changes to green.

 

Enables you to change the scale of the Alignment Plane. When some Respondents are very closely aligned, they can be made to move apart by zooming in.

 

The "spotlight" can be put on an Respondent once it is selected. This changes the Respondent's object to blue. The spotlight is applied by selecting the Respondent in the Selected Respondent list. The spotlight can be placed on more than one Respondent by holding the Ctrl key down while applying the spotlight.

 

Causes the analysis to disregard Respondents that do not have enough actual data to contribute to the analysis. By virtue of how alignment is calculated, Respondents with very little data might show up as being closely aligned. When you find that the primary reason Respondents are clustered together is by virtue of what the did not say, as opposed to what they said, you can eliminate them from the analysis by setting the Minimum Alignable Items value to a number greater than zero. Any Respondent that does not have content coded to at least this many Alignment Items is excluded from the analysis. Excluding them can be important, not only because their proximity to each other is meaningless, but also because their presence influences the plotting of all other Respondents.

 

This setting determines how closely Respondents must be aligned around an Alignment Item in order to show up in the Basis for Alignment list. By default, an item appears in the Basis for Alignment list only when all of the Respondents in the Selected Respondents list share this item. By lowering the threshold from 100%, you can get an indication of what the Selected Respondents have in common as a group, but without forcing 100% alignment around individual items.

 

This setting determines at which level within your coding structure the analysis should be performed. An option exists for each Code Level, and they are arranged from lowest to highest level. In general, you should perform the analysis at the deepest level, because it produces the maximum number of codes with which to differentiate Respondents.

 

There are a few factors that determine your selection here. Below are two examples to give you an idea of why you might choose one level or another.

 

Example 1: You want to see how Respondents align around a specific category, for example Strategy.

Check Strategy in your Category list, and select Sub-Issue or Issue for Analysis, depending on which level you code your comments to.

 

 

Example 2: You use a standard interview protocol where you asks every Respondent questions around various categories (e.g., Strategy, Organization Structure, and Management Model), and you want to get an indication of how thorough a job your team has done collecting data.

In this case, you should have data in every category for every Respondent. You can test this by setting the Alignment Level to the top level (Category by default) and run the analysis with no filters. If all Respondents appear to be 100% aligned, it means you can collected data in all categories. Any differences among the Respondents is the result of missing data.

 

Performs the analysis using the current settings.

 

The Distance for a Respondent is the sum of the Alignable Items that the Respondent does NOT have in common with the other Respondents. Overall, the Respondent with the most in common with the other Respondents has the smallest number, while the Respondent with the largest number has the least in common with the others. (This can be observed graphically by putting the spotlight on specific Respondents.)

 

This list contains each Alignment Item that the Respondents in the Selected Respondents list have in common. By default, an item is added to this list only if all of the Selected Respondents have the item in common. This constraint can be relaxed by changing the Popularity Threshold to less than 100%.

 

Output from an analysis can be accessed and exported to other tools through the Tools menu. The data is copied to the clipboard and can be pasted directly into a spreadsheet such as Excel. The export options are as follows:

 

 

XY Scatter Values

This option exports the geographic coordinates produced from the Alignment Analysis, along with the names and Demographic values for each of the Respondents. A chart can be created from this data in Excel by selecting XY Scatter in Excel’s chart wizard as follows:

 

1. After running the analysis, click on Tools/Export/XY Scatter Values. The values will be copied to your computer's clipboard.

 

2. Open Excel and paste the clipboard contents into a cell.

 

3. Select the cells from the columns labeled X and Y.

 

4. Open the Excel Chart Wizard, select XY Scatter-Gram, then click finish.

 

 

 

MDS Matrix

This option exports the half-matrix distance values used in multi-dimensional scaling. Each value represents the distance between a pair of Respondents. The distance represents the number of Analysis Items (usually Issues or Sub-Issues, depending on the level at which the analysis is performed) that the pair of Respondents do not have in common. This data can be imported into a statistics package (such as SPSS) for further analysis.

 

 

This list contains the Codes that were used in the analysis. The items in this list are determined by the Alignment Level, in conjunction with any filters you set. The Popularity column is the percent of Respondents with content coded to the Code or one of its descendents.

 

When an item in this list is selected, all the Respondents with content coded to this item are selected in the Alignment plane. Multiple items can be selected by holding the Ctrl key down while selecting.

 

This is a measure of how well the Respondents can be plotted on a two-dimensional plane. Given the sum of the distances between all the Respondents, Stress is the portion of that sum that remains unaccounted for after the two-dimensional best fit is found, and is represented as a percent of the sum of the distances. Plotting in more than two dimensions produces less stress.

 

If the Residual Stress value is high, it doesn't mean the analysis is invalid, although one should consider completeness of information, consistency in coding, and other factors that can lead to inaccurate conclusions. It is possible that most of the stress is due to a two-dimensional representation of object relationships that can only be explained in three or more dimensions. The risk of relying on a scatter-gram with high stress is that it may not provide an accurate representation of the relative distances between the Respondents. However, the distance calculations (in the Selected Respondent List and the Export options) are always accurate based on the data you have collected and coded, and can be relied upon to the extent that you trust your data.

 

 

Clicking this button toggles the Filter Group in and out of view.

 

By default, results are displayed only for Respondents with Comments that fall in the scope of the analysis. When the Include Empty Results checkbox is checked, output is produced for all Respondents. This is useful when you are interested not only in seeing who is aligned around a theme, but also who did not express a view toward a particular theme at all. Respondents that match this description will appear clustered together in the scatter gram, away from those that expressed views.