This document describes establishing a calibration model between sensory properties and consumer preference of raspberry jam. It involves relating sensory variables from jam samples to consumer preference rankings through partial least squares (PLS) regression models. Two PLS models will be built - one using principal component regression (PCR) and another using partial least squares (PSR). The models will be compared to determine which has the best predictive ability and relationship between sensory variables and consumer preference.
This document describes establishing a calibration model between sensory properties and consumer preference of raspberry jam. It involves relating sensory variables from jam samples to consumer preference rankings through partial least squares (PLS) regression models. Two PLS models will be built - one using principal component regression (PCR) and another using partial least squares (PSR). The models will be compared to determine which has the best predictive ability and relationship between sensory variables and consumer preference.
This document describes establishing a calibration model between sensory properties and consumer preference of raspberry jam. It involves relating sensory variables from jam samples to consumer preference rankings through partial least squares (PLS) regression models. Two PLS models will be built - one using principal component regression (PCR) and another using partial least squares (PSR). The models will be compared to determine which has the best predictive ability and relationship between sensory variables and consumer preference.
This document describes establishing a calibration model between sensory properties and consumer preference of raspberry jam. It involves relating sensory variables from jam samples to consumer preference rankings through partial least squares (PLS) regression models. Two PLS models will be built - one using principal component regression (PCR) and another using partial least squares (PSR). The models will be compared to determine which has the best predictive ability and relationship between sensory variables and consumer preference.
properties and consumer preference of a raspberry jam.
Problem: we want to assess jam quality and we would like to know which sensory properties consumers prefer to find when choosing a jam. To do that we will relate the sensory variables of a set of jam samples with the consumer preference through a calibration model.
Filename: JAM.
1. Data set. The data set contains a training set of ! jam samples. The samples have been picked up at four different places "#$#%& and harvested at three different times "'$'(&. Sample name Place Harvesting time #)' #!)' ! #()' ( #%)' % #)'! ! #!)'! ! ! #()'! ( ! #%)'! % ! #)'( ( #!)'( ! ( #()'( ( ( #()'% % (
*e have three different data sets related to these samples+ the ,sensory-. which contains sensory variables provided by e/perts0 the ,instrumental-. which contains instrumental measurements about the jam samples and the ,1reference-. which ranks the preference of the consumers. 2ensory variables and preference have been estimated on a scale going from 3 to 3.
The description of the instrumental variables is as follows+
Nr. Name Description 4 2pectrophotometric colour measurement ! a 2pectrophotometric colour measurement ( b 2pectrophotometric colour measurement % Absorb Absorbance 5 2oluble 2oluble solids "6& 7 Acidity Titrable acidity "6&
And the description for the sensory variables is below+
Steps 1. Do te P!" calibration mo#el ta$ing % &sensory' an# ( preference'. a. )as$ * "egression * P!" 2ample set+ Training. B$variables+ 2ensory. +eigts: All @2dev. ,ali#ation meto#: #ross ?alidation. 2etup+ 9ull #ross ?alidation. !enter #ata "selected&. C$variables+ 1reference. +eigts: All @2dev. ,ali#ation meto#: #ross ?alidation. 2etup+ 9ull #ross ?alidation. !enter #ata "selected&. -oo$ at te si.e of te mo#el. Plot of ( resi#ual vali#ation variance /bottom left0. The minimum indicates the siDe of the model "(&. Plot ypre# vs. ymeas 1it te correct si.e of te mo#el. Plot2Pre#icte# vs. 3easure#2Suggeste# number of components "(&. Eo we have a good predictive abilityF -oo$ at te variance explaine# in % an# (. *hy does the first 1# e/plain so little variance of CF -oo$ at te score plot. #an you see patterns in the samples according to the place where the raspberries were picked upF And according to the harvesting timeF -oo$ at te loa#ings plot. *hich variables of % "sensory& are correlated among themF Eirectly. inverselyF Got correlated at allF *hich variables of % correlate best to (F *hich sensory properties prefer the consumers in a jamF
4. Do te P-S calibration mo#el ta$ing % &sensory' an# ( preference'. b. )as$ * "egression * P-S1 2ample set+ Training. B$variables+ 2ensory. +eigts: All @2dev. ,ali#ation meto#: #ross ?alidation. 2etup+ 9ull #ross ?alidation. !enter #ata "selected&. C$variables+ 1reference. +eigts: All @2dev. ,ali#ation meto#: #ross ?alidation. 2etup+ 9ull #ross ?alidation. !enter #ata "selected&. -oo$ at te si.e of te mo#el. Plot of ( resi#ual vali#ation variance /bottom left0. The minimum indicates the siDe of the model "!&. Plot ypre# vs. ymeas 1it te correct si.e of te mo#el. Plot2Pre#icte# vs. 3easure#2Suggeste# number of components "!&. Eo we have a good predictive abilityF -oo$ at te variance explaine# in % an# (. *hy does the first 1# e/plain already a lot of ( varianceF -oo$ at te score plot. #an you see patterns in the samples according to the place where the raspberries were picked upF And according to the harvesting timeF -oo$ at te 1eigts plot. *hich variables of % "sensory& are correlated among themF Eirectly. inverselyF Got correlated at allF *hich variables of % correlate best to (F *hich sensory properties prefer the consumers in a jamF
5. !ompare P!" an# P-S mo#els. a. Eo they have the same siDeF *hyF b. *hich has the best predictive abilityF