Get documents about "Application of PLS regression on binary qualitative data Issues
Document Sample
Application of PLS regression on binary qualitative data
Issues of taste performance, ageing, experience and
exposure
Janna Bitnes, Ph.D. Student
Matforsk, Norwegian Food Research Institute, Norway
Royal Veterinary and Agricultural University, Denmark
7th Sensometrics Meeting
July 28-30 2004
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 1 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Background
Taste impairment with age
Learning effect with sensory experience and exposure
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 2 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Are they too old?
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 3 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Research question
Can ageing, experience and exposure explain the variation in taste
performance of the judges of the sensory panel?
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 4 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Subjects
The sensory panel at
Matforsk
10-12 members
Specifically selected,
employed and educated
as sensory panellists
(ISO standards)
Perform sensory work on
all kinds of food products
three days a week
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 5 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
39
38
37
36
35
34
33
32
31
30
29
28
27
Participation in the
26
25
24
23
22
basic taste identification
Judge #
21
20
19
18
test during the period 17
16
15
14
1976-2003
13
12
11
10
9
8
7
6
5
4
3
2
1
30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66
Age (year)
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 6 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Stimuli S a m p le C o n c e n tra tio n T a s te s tim u lu s
1 0 ,2 % S u c ro s e
2 0 ,4 % S u c ro s e
13 different samples 3 0 ,6 % S u c ro s e
4 0 ,0 1 % S o d iu m C h lo rid e
Sucrose, sodium chloride, 5 0 ,0 3 % S o d iu m C h lo rid e
citric acid and caffeine 6 0 ,0 6 % S o d iu m C h lo rid e
dissolved in water 7
8
0 ,0 0 5 %
0 ,0 1 %
C itric a c id
C itric a c id
Three different 9 0 ,0 1 5 % C itric a c id
10 0 ,0 6 ’1 0 -3 % C a ffe in e
concentrations 11 0 ,1 4 ’1 0 -3 % C a ffe in e
To provide the basic taste 12
13
0 ,2 7 ’1 0 -3 %
100 %
C a ffe in e
W a te r
stimuli of sweet, salt, sour
and bitter plus water
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 7 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Identification test
13 Samples (25 ml) were presented with a three-digit random number
Random order within each test session and identical for all subjects.
There were two replicates divided in two sessions one before and one 1 hour
after lunch.
The sensory panellists responded with one of the five choices: sweet, salt,
sour, bitter or water.
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 8 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Data analysis
Partial Least Squares Regression (PLSR): To what extent the
background variables ageing, experience and exposure explain the
variation in the responses from the judges.
3 13
Age Experience Exposure Responses
X Y
17 217
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 9 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
PLSR, X variables
X = Ageing, experience and exposure
Ageing: measures the specific age of the
panellists at the time of the test
Experience: measures number of years the judges
have participated in the sensory panel
Exposure: measures number of performed
identification tests for the individual panellists
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 10 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
PLSR, Y variables
Y = Responses, taste performance measured as mean of correct
identifications of sweet, salt, sour, bitter and water
Responses corrected
Ageing, experience and exposure SweL SweM…………….Water 13
X Y
217 217
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 11 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 12 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
A condensed way to show significance and sign of effect
-log10(p(bkj)) * sign(bkj)
The significance levels
P(bkj) were transformed by
-log(p) so that high
numbers represent high
Y-Variables
reliability.
For combining interpretation of the
main effects and their reliability, the
sign of the main effects was
combined with the transformed
significance levels
X-Variables
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 13 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Matrix plot of -log10(p(bkj)) * sign(bkj) for PC2
-0.494 9.401e-02 0.682
* 1.271
** 1.859
*** 2.447
SweetL
SaltL
SourL
BitterL
Y-Variables
SweetM
SaltM
SourM
BitterM
SweetH
SaltH
SourH
BitterH
Water
Ageing Experience Exposure
ISBAPLSRJK2PCsSorted - Matrix Plot, Sam.Set: All Samples, Var.Set: Selected Variables
X-Variables
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 14 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Conclusion - are they too old?
No!
There might be objections to use older subjects without
previous sensory experience as sensory judges, but the general
tendency is that when the panellists first are experienced with
sensory work there is no major taste impairment with ageing.
Thank you for your attention ;-)
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 15 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 16 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
APLSR was performed in two steps
Indicator matrix 1/0 Responses
Step 1: Judge correction
X Y
Age, experience.. Responses corrected
Step 2: Effect of age and experience
X Y
ember of the Food Science Alliance : NLH - Matforsk - Akvaforsk Sheet nr.: 17 | 8/12/2004 | Author: .... | Archive: I:\felles\_Utve
Related docs
