***** SIMS: Penalty Analysis - Basic Usage Notes *****
Purpose:
--------
Penalty analysis is a method used in sensory data analysis to identify potential directions for
the improvement of products, i.e. Overall Liking, on the basis of the other sensory attributes
presented to consumers or experts.
Penalty importance is measured by comparing the Overall Liking ratings of those consumers who
thought the product was Too Much (TM) or Not Enough (NE) on a particular JAR attribute with
the ratings of those consumers indicating that the JAR attribute was Just About Right (JAR).
The sensory test for Penalty Analysis must include:
a) An Overall Liking attribute for the global satisfaction index for a product/sample.
Typically this Overall Liking attribute would be a standard 9 point scale.
b) One or more JAR attributes for individual characteristics of the product.
Typically these JAR scales are a standard 5 point scale.
Example: Way Too Little ; Too Little ; Just About Right ; Too Much ; Way to Much
Example: Much Too Weak ; Somewhat Too Weak ; Just About Right ; Somewhat Too Strong ; Much Too Strong
JAR question choices should be, LEFT to RIGHT: Much Too Weak ; Somewhat Too Weak ; Just About Right ; Somewhat Too Strong ; Much Too Strong
or if vertical scales, TOP to BOTTOM: Much Too Weak ; Somewhat Too Weak ; Just About Right ; Somewhat Too Strong ; Much Too Strong
Excel reports JAR Distribution Tab assumes NE (Not Enough) is LEFT of center JAR, and TM (Too Much) is RIGHT of center JAR.
Glossary:
---------
JAR = Just About Right
NE = Not Enough
TM = Too Much
N = Number of Panelists
Interpretation:
---------------
Penalty analysis shows the amount that the Overall Liking was penalized by the Not-JAR respondents.
The Overall Liking score would increase by the Total Penalty.
The JARs with the larger values for Total Penalty may be of interest.
The graph values, units of measure, will depend on questionnaire design and panelist's data.
Generally speaking, with a typical Overall Liking 9pt scale, 1 to 9 return values,
a JAR with a total penalty > 0.50 is high and > 0.25 is noteworthy.
Interpretation is the responsibility of the Sensory professional taking all aspects into account.
When to Use Penalty Analysis:
-----------------------------
- Studies with Overall Liking and JAR scales.
- Monadic or multi-sample studies. One or more samples.
- More valuable with larger test populations (N).
- More valuable when parallel description sensory data is available.
Qualifying Attributes:
----------------------
Overall Liking, SIMS will *try* to automatically identify your Overall Liking Scale
by comparing the attribute's data export label with common ones, such as 'Overall Liking'
You may need to manually select the proper attribute for your ballots 'Overall Liking', see 2nd Tab.
JAR scales must be Hedonic attribute types. Line Scale types are not supported.
JAR scales must have an ODD number categories. The middle category is assumed to be JAR.
JAR scales actual Not-JAR responses must be >= 20% to be included, in either direction, NE or TM.
JAR scales Mean-drop calculation must be > 0 to be included.
FAQ: Where is my JAR question?
Hint: Is your JAR question a *Hedonic* question type in SIMS? Line Scale types are not supported.
Hint: JAR scales actual Not-JAR responses must be >= 20% to be included, in either direction, NE or TM.
Reference: Tom Carr's PowerPoint document: Penalty Analysis Using JAR Scales, Penalty_Analysis_Overview.ppt
Options:
----------------------
[X] Only Show JARs
Optional, selection list will filter and only show JAR detected hedonics,
having either a) Data Export Label 'JAR' or 'Just *', or b) middle choice 'JAR' or 'Just *'
JAR Distribution Graph:
(*) Graph the N Counts
( ) Graph the %N Counts
This is the JAR Distribution Graph on the JAR Distribution tab in Excel. A stacked bar graph. N Counts are more common.
[X] Rare, override Qualifying Attributes 20% rule to 5%
Rarely used option. Lowers the JAR scales Not-JAR responses to only 5% instead of the normal >= 20% to be included on the report.
Mathematics Information:
------------------------
SIMS penalty analysis calculations are simple mathematics. SAS software is not required.
The mathematic specifications per Tom Carr, 2007.
The qualified and selected JAR scales are independently analyzed with overall liking.
Collapse the JAR scale responses by --Panelists-- into three, NE, JAR, and TM.
Calculate the % of panelists for the JAR scale for the two non-JAR groups, NE and TM.
JAR(NE%) = N of JAR's(NE) / Total JAR's N
JAR(TM%) = N of JAR's(TM) / Total JAR's N
If JAR(NE%) >= 20% (i.e. If at least 20 percent of panelists answered NE)
Mean Drop = Average Overall Liking(only using JAR's JAR panelists.)
- Average Overall Liking(only using JAR's NE panelists.)
Total Penalty = Mean Drop X JAR(NE%)
If JAR(TM%) >= 20% (i.e. If at least 20 percent of panelists answered TM)
Mean Drop = Average Overall Liking(only using JAR's JAR panelists.)
- Average Overall Liking(only using JAR's TM panelists.)
Total Penalty = Mean Drop X JAR(TM%)
Other related reading materials:
--------------------------------
Tom Carr's PowerPoint document: Penalty Analysis Using JAR Scales, Penalty_Analysis_Overview.ppt
Google search on the Internet. Try searching for: Sensory "Penalty Analysis"
SIMS Examples and related reading materials:
--------------------------------------------
http://www.SIMS2000.com/ReportsShowPenaltyAnalysis.asp
|