Linearity,

linear range, sensi.vity

Linearity

Methods ability to obtain signals

which are directly propor.onal to the
concentra1on of analyte in the sample.

1
 Linear range
Range of concentra.ons where the
signals are directly propor1onal to the
concentra1on of the analyte
in the sample.

Sensi.vity
The change in instrument response
which corresponds to a change in
the measured quan1ty.

The gradient of the response curve.

2
 Experiment planning
for evalua.on of linearity
•  Type of calibra1on samples
•  Concentra1on range
•  Number of concentra1on levels
•  Measurement protocol

Type of calibra.on samples

•  Standard solu1ons
–  matrix-free

•  Matrix-containing samples
–  Matrix-matched calibra1on
–  Blank matrix extract (preferably the same type as the sample)

3
 Concentra.on range
•  Appropriate for the method
•  Expected working range ± 10%…20%
•  70%…130% of expected analyte concentra1on
•  Blank sample

Number of concentra.on levels

•  Minimum 6 concentra1on levels are acceptable
–  Suggested 10
–  Evenly placed
–  Prepared from independent dilu1ons

4
 Measurement protocol
•  As similar as possible to the real life situa1on
–  Random order
–  Between the samples

•  Analysed at least twice

Evalua.on of linearity
•  Visual evalua.on

•  Residuals

5
 Visual evalua.on
120000000 80000000
LINEAR 70000000 NOT LINEAR
100000000
60000000
80000000
50000000
Signal (au)

Signal (au)
60000000 40000000

30000000
40000000
20000000
20000000
10000000

0 0
0 2 4 6 8 10 0 1 2 3 4 5 6
c (mg/kg) c (mg/kg)

Residual
•  Difference between experimental signal and calculated signal
yi ŷi
120000000

100000000
yi
80000000
Signal (au)

60000000

40000000 ŷi = slope ⋅ xi + intercept

20000000

0
0 2 4 6 8 10 εi = yi − ŷi
c (mg/kg)

6
 Visual evalua.on
120000000 140000000
80000000
LINEAR 70000000
120000000
NOT LINEAR
LINEAR ?
100000000
60000000
100000000
80000000

Signal (au)
Signal (au)

50000000
80000000

Signal
60000000 40000000
60000000
30000000
40000000
40000000
20000000
20000000 20000000
10000000

0 0 0
0 2 4 6 8 10 0 0 1 2 2 4 6
3 8
4 10
5 12
6
c (mg/kg) c (mg/kg)
c (mg/kg)

Plot of residuals

7
 Linear, homoscedas.c
10000000

8000000

6000000
Residual (au)

4000000

2000000

0
0 0.2 0.4 0.6 0.8 1 1.2
-2000000

-4000000

-6000000
c (mg/kg)

Linear, heteroscedas.c
30000000

20000000

10000000
Residual (au)

0
0 2 4 6 8 10 12
-10000000

-20000000

-30000000
c (mg/kg)

8
 Not linear, heteroscedas.c
30000000

20000000

10000000
Residual (au)

0
0 2 4 6 8 10 12
-10000000

-20000000

-30000000

-40000000
c (mg/kg)

Rela.ve residuals

•  ±20% is acceptable
yi − ŷi
Yi =
ŷi

9
 Correla.on coefficient r
Coefficient of determina.on r2
•  Total variability in the response that is
accounted for by the model

•  Not a measure of linearity!

Linear range

10
 “range”

Dynamic range, Analy.cal range

120000000

100000000
Signal (au)

80000000

60000000

40000000

20000000

0
0 2 4 6 8 10 12
c (mg/kg)

11
 Dynamic range, Analy1cal range

Working range, Calibra.on range
120000000

100000000
Signal (au)

80000000

60000000

40000000
*LOQ
20000000

0
0 2 4 6 8 10 12
c (mg/kg)

Dynamic range, Analy1cal range

Working range, Calibra1on range
120000000

Linear range,
100000000 Linear dynamic range

Signal (au)

80000000

60000000

40000000
*LOQ
20000000

0
0 2 4 6 8 10 12
c (mg/kg)

12
 Sensi.vity
•  gradient of the calibra1on graph

Sensi.vity
120,000,000 ΔSignal
Slope =
100,000,000 Signal = 7610033⋅ c + 32270418 Δc
80,000,000
Signal (au)

ΔSignal
60,000,000

40,000,000

20,000,000 Δc
0
0 2 4 6 8 10
c (mg/kg)

13
 Sensi.vity
•  Method op1miza1on

•  Quality assurance

•  Rou1ne monitoring of the instrument

Connected to other parameters

•  Calcula1ng matrix effect Slopematrix−matched
%ME = ⋅100%
calibra1on graph method Slopesolvent

•  Calcula1ng LoD and LoQ Sy. x
LoD = 3.3×
Slope

Sy. x
LoQ = 10 ×
Slope

14