A number of students have expressed interest in using additional features of Minitab to permit them to dig more deeply into the Chicago Fashion data set. This memo explains two techniques that could be helpful in this regard:   Three-Way Tables and Working with a Subset of the Data.

1.  Three way tables.

Once you have made a two-way table, you can add the MEAN of a third variable by simply clicking on "Summaries" to get a new dialog box, then selecting the variable of choice, and then checking the box for "Means," as shown below.

For example, you might look at a table of low to high income shoppers (C47) who also shop less (=1) to more (=5) at LRD-TYLR (C54). Notice from the dialog box below that for this table, after selecting the variables, we checked on Column percents, Chi-square analysis and Show Count.

For this table, you can then click on "Summaries" to get a new dialog box, then select DESIGNRS (C1), and then check the box for "Means." This way get mean scores for each group on the variable "I frequently buy clothes and accessories made by designers" (the variable coded into C1 DESIGNRS).

The result is a table of counts and means for each cell. To see the actual table showing "Rows: INCOME Columns: LRD&TYLR" click on: Link to Table 1 (See table 1)

 
This table shows that heavy Lord & Taylor shoppers are more likely to say they buy designer products, since the means in the very bottom row (the "All" row) increase as you read across the table. The table also shows that higher income shoppers do the same, because the means in the rightmost column (the "All" column) increase as you read down the table.

2 Working with a subset of the Data.

There is also a relatively easy way to extract a sub-set of the data and place it in a different location in the "Gurney" file – and then use it to explore a sub-set of customers in more depth.

For example, suppose you are Lord & Taylor and you want to attract good Lane Bryant shoppers. So you want to know more about the "4s and 5s" on Lane Bryant (C53). Specifically, you want to know about how much they already shop Lord and Taylor (C54), their incomes (C47), and their Designer (c1) scores. So you can make a special database with just these people and these variables – and then use it as a regular Minitab database within "Gurney".

It takes just three steps using the "Manip" screens:

• Click on Manip/Copy Columns to get the Copy Columns screen:

• Now select the four columns you want, and set the columns into which they are to be copied. Be sure to chose exactly the right number of columns.

Select the variable – Income – and the values 5:7 -- that you want – and click OK. You will have a new data base like this sample. Note that in the Income column (Column 80) there are only 5s, 6s and 7s.

 You can now use this database to make "Tables" or "Charts" as usual. For example, cross-tabulating "Lord & Taylor" (78) by "Lane Bryant" (C79), with "DESIGNRS" scores (C77) shown as means, you will get a table shwing how interest in DESIGNERS varies by shopping behavior. Click to Link to Table 2. (See Table 2).

In this Table:

• Notice that you are now working with a much smaller sample – just 754 high income shoppers– so for this example many table cells are empty and chi-square may warn you that relationships are questionable. But this technique may be very helpful for digging more deeply into the data.
•  In this instance, the data show that:
• Most of the sub-sample are heavy Lord and Taylor shoppers, because they fall into the top rows.
• Most of the sub-sample are also heavy Lane Bryant shoppers, because they fall into the first couple of columns.
• Within this pattern, we also see that low "DESIGNRS" scores – meaning HIGH purchase of designer products – is generally most common among groups closest to the upper left – that is among heavy shoppers of both stores.