Purpose
The product characterisation in the suite of
sensory analysis can help identify the most and least characterizing attributes
for each product, or group of products and demonstrates their relative
discriminating ability.
Note that for EyeOpenR to read your dataset, the
first five columns must be: Assessor, Product, Session, Replica
and Sequence. For more information on data format, see the sensory demo
data sets.
Background
The tool uses analysis of variance to check for significant
differences above and below the mean value of each attribute. For each attribute an ANOVA model is applied to
check if the scores given by the assessors are significantly different for the
product in question. The character of the product, displayed in terms of
attributes, is then visualised on a plot, with more discriminating attributes
highlighted.
Options
The options are very simple to setup and
relate to output display only:
- Significance level (5%, 10%, 20%) – the user can choose what level
of statistical significance (5%, 10% or 20%) the software marks as
‘significant’ using colours and for the extra V-test charts.
- No. of decimals for Values – number of decimal places displayed in
the adjusted means and v-test outputs
- No. of decimals for P-values– number of decimal places displayed for
any p-values in the output
Results and Interpretation
- Adjusted Means - the first
tab reports the values of the adjusted means for each sample and attribute
(derived from the ANOVA model fitted) both in a table and then below as a plot.
In the table, for each product, attributes with a score statistically
significantly higher than the mean for that attribute are coloured blue,
whereas those coloured red are statistically significantly lower. Cells that
are white are indicating products that are not statistically significant
different from the mean attribute score for the attribute. Below the table is a line
chart showing the same scores with one line per product across each of the
different attributes.
- V-test Plot – the V-test plots are displayed on one tab for each
product. The plot shows the size of the V-test score for all the attributes and
ties in with the previous table of adjusted means. Where a product
has a lower adjusted mean score for an attribute when comparing to the
attribute mean, the v-test score will be negative and the plot shows this in
red (to the left). Where a product
has a higher adjusted mean score for an attribute when comparing to the
attribute mean, the v-test score will be positive and the plot shows this in yellow
(to the right). The user can
quickly see which attributes are scoring higher and lower than average for the
product. The V-test statistics
are shown in the table below the V-test plot, with the associated mean scores
and p-values for all the attributes for that product. The V-test score can be
understood to be similar to the normal Z-statistic, as a quantile of the normal
distribution and is a transformation of the p-value, i.e. when the absolute
value of the V-test statistic is > 1.96 then the p-value will be
statistically significant at p=0.05. The V-test statistic helps by showing
whether the difference is positive or negative.
- V-test Plot (only signif.) – This plot and table are the same as the
previous, however this time only those attributes that are statistically
significantly different at the selected significance level are shown on the
plot and in the table.
- R packages used – SensoMineR,
FactoMineR
References
Husson F.
, Lê S. and Pagès J. (2009). SensoMineR dans Evaluation sensorielle - Manuel
méthodologique. Lavoisier,
SSHA, 3ème édition.
Lê S. and
Husson F. (2008).
SensoMineR: a package for sensory data analysis. Journal of Sensory Studies. 23(1). 14-25.