Some formatting depends on the type of model you estimated, but some is common to all types. This page covers the latter. For model-specific formatting, click on one of the following links:
Cluster model
DFactor model
Choice model
Regression model
We will use the Latent GOLD shoe example to illustrate formatting that applies to all models. Specifically, we will cover these worksheets:
Overall summary
Summary
Parameters
Profile
ProbMeans
The Wizard has cleaned this sheet up by adjusting column widths, but it has done more.
It has selected the best model, using lowest BIC as a criterion; it has highlighted this model in row 7; and it has placed a BIC chart in the worksheet. It has also changed the color of the tabs to green, signifying that this is the model with the lowest BIC.
The summary sheet has been cleaned up in terms of adjusting column widths. It has also turned the diagonal of the Classification Table and Prediction Table boldface.
Off to the right of the Prediction Table, it has constructed a row-percentage version of the same data.
The only other change has been the additions to cells E1:F3, where the Wizard placed some housekeeping items.
This sheet is where extensive formatting begins. Notice that many cells are colored orange, green and yellow.
A key to this color scheme appears in cells L8:L10. Orange cells denote the largest parameter value in a row; yellow green, the lowest value; and khaki, higher than average. In practice this means that for those parameters where increased value means increased utility, orange cells identify the segment where the utility has particular emphasis. Often this provides clues to segment characteristics and can be helpful in naming the segments. We see in the example above that Class1 has a strong preference for modern fashion (cell B8) and standard quality (cell B12). The low value for the None parameter in cell B16 suggests they have a high purchase rate. We discuss the interpretation of this particular model in more detail in the choice model tutorial. Here we just want to introduce the color-coding scheme.
All p-values in the above example are significant, so no special formatting appears in columns F or H, However, we can see what happens when we encounter a p-value less than 0.05 by overriding model results and typing 0.1 in cell H8. When you do this, the cell turns boldface red, as you can see below.
Note that placeholder cells B1:D1 are selected by default. This is where you enter segment names once you feel you can properly characterize them. Each of the tutorials discuss this in more detail.
The same color-coding scheme just described applies to the profile sheet as well, except comparisons are made vertically between attribute levels. Corresponding to what we found in the parameters sheet, we see in cells B5:B6 that Class1 has a much higher conditional probability of choosing a shoe with a modern rather than traditional fashion. We also see in cells B8:B9 that the preference for standard quality shoes is much less pronounced.
Interpretations of formatting for covariates is similar. Members of Class1 are likely to be young (cells B26:B28) and female (cells B23:B24).
You can see this even more clearly from the charts that the Format Wizard has placed to the right of the table. Examining these charts, we see that segment 1consists largely of young females with a strong preference for modern shoes. By contrast, segment 2 shows a strong preference for quality from an older clientele. Segment 3, more than 86% male, prefers high-quality and modern style, though not to the intense degree exhibited by segments 1 and 2.
The conditional formatting and charts greatly facilitate interpreting results from the Profile sheet.
Moving on to the ProbMeans sheet, we find the same color formatting, except this time the comparison is across rows rather than columns.
Classes are always presented in decreasing size, and in this case Class1 is much larger than the other two. As a result, most of the cells in column B are orange. That makes the orange cells in columns C and D all the more interesting. We see from cells C6 and C26 that members of Class2 tend to prefer traditional shoes and fall into older age cohorts.
Formatting for these sheets appear in all model output, but each model type contains its own special formatting. For a tutorial on a specific model type, click one of the links below.