Microsoft Office Tutorials and References

In Depth Information

For the moment, let’s assume that the random sample has selected a proportion-

ally fair representation of respondents, and this is precisely what TiendaMía.com

desires. Thus, 28% (8

÷

29, 8 of 1

\

1 respondents out of 29) should be relatively

close to the population of all 1

1’s in TiendaMía.com’s customer population. If we

want to account for the difference in respondent category size in our analysis, then

we will want to calculate a
weighted
average of favorable ratings, which reﬂects the

relative size of the respondent categories. Note that the ﬁrst average that we calcu-

lated is a special form of weighted average: one where all weights were assumed to

be equal. In range G25:G28 of Exhibit 5.24 we see the calculation of the weighted

average. Each average is multiplied by the fraction of respondents that it represents

of the total sample.
8
This approach provides a proportional emphasis on averages. If

a particular average is composed of many respondents, then it will receive a higher

weight; if an average is composed of fewer respondents, then it will receive a lower

weight.

So what do our respondent weighted averages (G25:G28) reveal about Products

compared to the equally weighted averages (F25:F28)? The results are approxi-

mately the same for
Products 1
and
4
. The exceptions are
Product 2
with a somewhat

stronger showing from 0.4965 to 0.5172 and Product 3 with a substantial drop in

score from 0.3073 to 0.2931. Still, there is no change in the ranking of the products;

it remains P-1, P-2, P-3, and P-4.

What has led to the increase in the
Product 2
score? Categories 1

\

2

are the highest favorable ratings for
Product 2
; they also happen to be the largest

weighted categories (8/29

\

1 and 2

\

0.310). Larger weights applied to the

highest scores will of course yield a higher weighted average. If TiendaMía.com

wants to focus attention on these market segments, then a weighted average may

be appropriate.
Market segmentation
is in fact a very important element in their

marketing strategy.

There may be other ways to weight the favorable ratings. For example, there

may be categories that are more important than others due to their higher spending

per transaction or more frequent transactions at the site. So, as you can see, many

weighting schemes are possible.

=

0.276 and 9/29

=

5.6 Summary

Cross-tabulation analysis through the use
PivotTables
and
PivotCharts
is a simple

and effective way to analyze qualitative data, but to insure fair and accurate analysis

the data must be carefully examined and prepared. Rarely is a data set of signiﬁcant

size exempt from errors. Although most errors are usually accidental, there may be

some that are intentional. Excel provides many logical cell functions to determine

if data have been accurately captured and ﬁt the speciﬁcations that an analyst has

imposed.

8
(0.7500
∗
8 + 0.5000
∗
6 + 0.6667
∗
6 + 0.5556
∗
9) / 29

=

0.6207.

Search JabSto ::

Custom Search