Вы находитесь на странице: 1из 12

Analytical Procedure comparison:

Auto Vs Manual titration for Salt


analysis
Background
• Auto titrator was installed at facility and various reading were taken for the samples

• Same samples were analysed manually and data was used for comparison of variations in

results.

• Total 196 readings were taken to compare the output at different salt concentration levels

Assumption made in statistical Analysis:

It was assumed that all the seasoning and FG was a part of same population from which

random sampling was done and analysed by two methods.

Thus the objective of analysis became: “To compare the variances of two results obtained

by two methods performed on the samples taken from same population”


Data Cleaning and segregation
• Three salt concentration ranges were created -

• Low Salt % - 1.8% to 2.8% (101 readings)

• Medium Salt % - 18.4% to 20.5% ( 65 readings)

• High Salt % - more than 20.5% (30 readings)

• Difference of Auto Vs Manual salt analysis across the different salt concentrations were
analysed
Analysis of Difference in Observations (Diff= Manual – Auto)
Hₒ: Difference in readings of auto Vs manual salt analysis is not Consistent
Difference in Observations (Manual -Auto)
62

A nderson-Darling N ormality Test


A -S quared 6.09
It is safe reject the null hypothesis
45 P -V alue < 0.005
M ean 0.04372
S tDev 0.26717 The difference across the
V ariance 0.07138 readings are normally distributed
S kew ness -0.10576 and is having a mean of 0.043
19 Kurtosis 3.24293 however some observations are
13 N 196 as high as -0.9 to 0.9%.
9 9 9 9 It is important to analyse the
6 M inimum -0.95000 variance of difference at different
2 2 2 2 2 2
0 1 0 1 1 1st Q uartile -0.04000 concentration levels.
-0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 M edian 0.06000
3rd Q uartile 0.13750
M aximum 1.01000
95% C onfidence Interv al for M ean
0.00609 0.08136
95% C onfidence Interv al for M edian
0.04000 0.07000
9 5 % C onfidence Inter vals
95% C onfidence Interv al for S tDev
Mean 0.24308 0.29660

Median

0.00 0.02 0.04 0.06 0.08


Paired T-Test and CI: Auto Titration, Manual Titration (For low
salt %)
Paired T for Auto Titration - Manual Titration

N Mean StDev SE Mean


Auto Titration 100 2.29010 0.31666 0.03167
Manual Titration 100 2.34620 0.31068 0.03107
Difference 100 -0.056100 0.073621 0.007362

95% CI for mean difference: (-0.070708, -0.041492)


T-Test of mean difference = 0 (vs not = 0): T-Value = -7.62 P-Value = 0.000

Patired T-Test results indicated that these a very high correlation between the auto
and manual titration at low salt% and the standard deviation in difference is 0.073
with std. error in mean to the tune of 0.007.

Regression analysis was further performed to establish the extent of correlation in


outputs of both the methods.
Régression Analysis: Auto vs Manual (Low Salt%)
The regression equation is
y = 0.0102x + 1.8299 Auto Titration = - 0.0358 + 0.991 Manual Titration
Regression Analysis- Low Salt readings R² = 0.9113

Predictor Coef SE Coef T P


Constant -0.03584 0.05661 -0.63 0.528
Manual Titration 0.99136 0.02392 41.44 0.000
2.81

S = 0.0739461 R-Sq = 94.6% R-Sq(adj) = 94.5%

2.61
Analysis of Variance

Source DF SS MS F P
2.41 Regression 1 9.3914 9.3914 1717.52 0.000
Residual Error 98 0.5359 0.0055
Total 99 9.9273

2.21
Unusual Observations

Manual Auto
Obs Titration Titration Fit SE Fit Residual St Resid
2.01
16 2.18 1.97000 2.12534 0.00840 -0.15534 -2.11R
23 1.89 1.99000 1.83784 0.01318 0.15216 2.09R
55 2.59 2.34000 2.53179 0.00942 -0.19179 -2.62R
74 2.68 2.42000 2.62102 0.01088 -0.20102 -2.75R
1.81
94 2.66 2.81000 2.60119 0.01054 0.20881 2.85R
1.96

2.01

2.16

2.35

2.42

2.80
1.81
1.83
1.92
1.94

1.97
1.99
1.99
1.99

2.01
2.03
2.06
2.09

2.29
2.31
2.33
2.34

2.36
2.40
2.41
2.42

2.43
2.43
2.78
2.79

2.80
2.81
2.82
2.84

100 2.74 2.84000 2.68050 0.01198 0.15950 2.19R

R denotes an observation with a large standardized residual.

Auto and manual results are highly corelated at low salt% (R-Sq=0.91)
Paired T-Test and CI: Auto Titration, Manual Titration (Medium
Salt%)
Paired T for Auto Titration - Manual Titration

N Mean StDev SE Mean


Auto Titration 64 19.6503 0.3890 0.0486
Manual Titration 64 19.6516 0.3543 0.0443
Difference 64 -0.001250 0.386833 0.048354

95% CI for mean difference: (-0.097878, 0.095378)


T-Test of mean difference = 0 (vs not = 0): T-Value = -0.03 P-Value = 0.979

Patired T-Test results indicated that standard deviation is up to 0.3% in the


observations at medium salt%. P- Value =0.979 (>0.005) suggest less possibility of
any statistical significance

Regression analysis was further performed to establish the extent of correlation in


outputs of both the methods.
Régression Analysis: Auto vs Manual Titration (Medium Salt%)
Correlation Curve for Manual Vs Auto Titration y = 0.4208x + 11.384
R² = 0.2134

20.4

20.2

20

19.8

19.6

19.4

19.2

19

18.8

18.6

18.4
18.40 18.90 19.40 19.90 20.40

Auto and manual results have no correlation at medium salt% (R-Sq=0.21)


Two-Sample T-Test and CI: Auto Vs Manual Titration(High Salt%)

Two-sample T for Auto Titration vs Manual Titration

N Mean StDev SE Mean


Auto Titration 32 27.783 0.406 0.072
Manual Titration 32 27.873 0.290 0.051

Difference = mu (Auto Titration) - mu (Manual Titration)


Estimate for difference: -0.090000
95% CI for difference: (-0.266774, 0.086774)
T-Test of difference = 0 (vs not =): T-Value = -1.02 P-Value = 0.312 DF = 56
Regression Analysis: Auto Titration versus Manual
Titration (High Salt%)
The regression equation is
Regression Analysis: High Salt Auto Titration = 6.62 + 0.759 Manual Titration

y = 0.3877x + 17.101
R² = 0.2945
Predictor Coef SE Coef T P
28.6 Constant 6.617 5.977 1.11 0.277
Manual Titration 0.7594 0.2144 3.54 0.001

28.4
S = 0.346627 R-Sq = 29.5% R-Sq(adj) = 27.1%

28.2
Analysis of Variance

28 Source DF SS MS F P
Regression 1 1.5069 1.5069 12.54 0.001
Residual Error 30 3.6045 0.1201
27.8 Total 31 5.1114

27.6 Unusual Observations

Manual Auto
27.4 Obs Titration Titration Fit SE Fit Residual St Resid
1 27.7 26.6500 27.6212 0.0764 -0.9712 -2.87R

27.2 R denotes an observation with a large standardized residual.


26.50 27.00 27.50 28.00 28.50 29.00

Auto and manual results have no correlation at high salt% (R-Sq=0.29)


Inference
• Auto titrator is effective and precise when used to analysed low salt%
• There will be a difference in reading to the tune of 0.3% when analysed
medium salt%
• Difference in analysis results can go as high as 0.9% when used to analyse
the high salt%

Recommendation:

It is recommended to use the diluted samples for analysis while using auto
titrator. It will improve the accuracy of the results.

% Error can be established by doing the repeated test with samples of known
concentration.
Test of repeatability: Z-score for triplet analysis
Auto Manual Z-Auto Z-Manual Based on available triplet analysis by both
19.65 20.12 -1.33779 0.87973 manual and auto method, Z-score confirms
19.64 20.12 -1.46622 0.87973 that readings for Auto is between (-1.4 to 1.1
19.65 19.94 -1.33779 0.37703 SD) and for manual; between (-1.3 to 1.1SD)
19.71 19.51 -0.56722 -0.82388
19.72 19.33 -0.43880 -1.32658 It is safe to assume that Auto and Manual
19.75 19.33 -0.05351 -1.32658 methods are repeatable with ± 1.4 SD.
19.81 20.20 0.71706 1.10316
19.80 20.20 0.58863 1.10316 SD_Auto=0.08
19.81 20.20 0.71706 1.10316 SD_Manual=0.36
19.84 19.45 1.10234 -0.99145
19.84 19.63 1.10234 -0.48874 It is confirmed that Auto titrator method is
19.83 19.63 0.97391 -0.48874 more repeatable.

It is recommended to produce more data at low salt% to validate the repetability at low salt% (Which is surly high)

Вам также может понравиться