MMPTM Proficiency Test 2009(2) · Product Research Division, Japan Tobacco Inc., JT Sumida Building, 17-7, Yokogawa 1-chome, Sumida-ku, Tokyo 130-8603, Japan JT International Germany

Coresta Sub Group

Monitoring and Maintenance of Physical Test Methods

Interlaboratory Proficiency Test 2009

Page 1 of 42

CORESTA Monitoring and Maintenance of Physical Test Methods Sub - Group

2009 Interlaboratory Proficiency Test

Issued by:

Graham Errington

R & D Centre

British American Tobacco

Regents Park Road

Millbrook

Southampton

United Kingdom

SO15 8TL

Phone: 02380 793540

e-mail: [email protected]

mailto:[email protected]

Coresta Sub Group



Page 2 of 42

Coresta Sub Group



Page 3 of 42

PARTICIPATING LABORATORIES British American Tobacco (Germany) GmbH, Weistrasse 26, 95448 Bayreuth, Germany

British American Tobacco, Denmark

Cerulean, Rockingham Drive, Linford Wood East, Milton Keynes, MK14 6LY, UK

Hauni

Imperial Tobacco - HH

Imperial Tobacco - LA

Product Research Division, Japan Tobacco Inc., JT Sumida Building, 17-7, Yokogawa 1-chome, Sumida-ku,

Tokyo 130-8603, Japan

JT International Germany GmbH, Zentrale Warenaufnahme, R & D Building/GQA-Laboratories,

Diedenhofener Str. 20, D-54294 Trier, Germany

Rhodia Acetow GmbH, Engesserstr. 8, D-79123 Freiburg, Germany

SODIM SA, 4 rue André Dessaux, 45404 Fleury les Aubrais cedex, France

ST Poland, 32-400 Myslenice, Jawornik 360, Poland

Papierfabrik Wattens GmbH & Co KG, Ludwig-Lassl-Strasse 15, A-6112 Wattens, Austria

Zhengzhou Tobacco Research Institute of CNTC, No. 2 Fengyang St., Zhengzhou New & High-Tech Industries, Development Zone, Henan 450001, China

Coresta Sub Group



Page 4 of 42

2009 CORESTA MMPTM SUB – GROUP INTERLABORATORY PROFICIENCY TEST

Introduction The purpose of an interlaboratory proficiency testing scheme

1 is to provide participating laboratories

with an objective means of assessing and demonstrating the reliability of the data they are producing, and is an important component of any laboratory quality assurance scheme. The importance of proficiency testing is recognised in ISO/IEC Guide 17025

2 where “participation in interlaboratory

comparison or proficiency testing programmes” is listed as an important component of quality assurance results.

Outline of Test Protocol Five filter types and five cigarette brands were selected to cover the range of interest for Individual Weight, Circumference, Pressure Drop and Ventilation.

The samples used in the study are presented in Table 1.

Table 1.

SAMPLE TYPE

SAMPLE ID VARIANT Pressure Drop

Type

Monoaceate

Filter

F1 Semi - Slim

F2 Slim

F3 King Size/ Regular High

F4 King Size/ Regular Low

F5 King Size/ Regular Medium

Cigarette

C1 King Size/ Regular

C2 King Size/ Regular

C3 Slim

C4 Super Slim

C5 Super Slim

Participating laboratories were instructed to measure a sub – group of 10 specimens under their normal operating conditions, repeated 3 times for each sample. Mean values of each replicate were reported in an Excel spreadsheet supplied with the protocol.

After completion, the spreadsheets were returned to the author for data analysis.

Data Analysis Statistical analysis was carried out using Microsoft Excel and MINITAB v15. All graphs were produced with MINITAB v15.

Coresta Sub Group



Page 5 of 42

Summary statistics were calculated for each sample, followed by individual value plots by laboratory, to identify potential outliers.

Precision Statistics

Following identification and removal of outliers, Reproducibility, R, and repeatability, r, were calculated for each sample and tabulated at the beginning of each section.

Mandel Plots

h statistics for between laboratory consistency, and k statistics for within laboratory consistency were calculated and plotted according to ISO 5725 – 2

3, for each product characteristic

Repeatability Variation (Cochran’s Test)

Cochran’s test4 was used to evaluate the within – laboratory performance of each laboratory

where smax is the highest standard deviation in a set of p laboratories. The following criteria were used to asses each laboratory:

If the test statistic, c, is:

- ≤ 5% critical value Accept

- > 5% critical value ‹ 10% critical value Straggler

- > 10% critical value Outlier

Z - scores

After removal of outliers, robust estimates of means of all laboratories were used as the reference for computing z – scores

Z – scores are plotted in ascending order. Laboratory performance is indicated using the following criteria:

|z| < 1 = good

1< |z| < 2 = satisfactory

2< |z| < 3 = questionable

|z| > 3 = highly questionable

Results are presented separately for Filters and cigarettes.

p

i

is

sc

1

2

2

max

outliers for adjusted deviation standard overall

value mean Total

value mean Single

-

s

X

x

s

XxZ

)(

Coresta Sub Group



Page 6 of 42

FILTER DATA ANALYSIS

Table 2. Filter Summary Statistics

Instrument Manufacturer/ Type Codes for Individual Value Plots

Instrument Manufacturer/Type Code

Borgwaldt 1

Cerulean 2

Sodiline 3

Sodimat 4

Sodimax 5

Sodimax CCD 6

VariableSample

IDN N* Mean StDev CoefVar Minimum Median

Maximu

mRange

F2 42 0 461.45 3.660 0.79 452.50 462.73 466.00 13.500

F1 41 1 510.45 3.770 0.74 501.30 512.00 518.00 16.700

F4 42 0 798.88 6.490 0.81 784.60 800.00 810.50 25.900

F5 42 0 894.37 6.400 0.72 878.40 895.23 903.00 24.600

F3 42 0 1044.20 8.000 0.77 1025.30 1047.00 1055.80 30.500

F2 42 0 18.00 0.028 0.16 17.93 18.01 18.04 0.117

F1 41 1 21.18 0.034 0.16 21.10 21.19 21.24 0.143

F5 42 0 24.47 0.037 0.15 24.39 24.47 24.55 0.165

F4 42 0 24.43 0.038 0.15 24.34 24.45 24.49 0.146

F3 42 0 24.43 0.049 0.20 24.23 24.45 24.49 0.260

F4 39 3 243.54 6.070 2.49 222.80 244.00 253.00 30.200

F1 38 4 396.98 12.760 3.21 351.90 396.60 414.30 62.400

F2 36 6 412.96 13.300 3.22 370.90 416.00 429.00 58.100

F5 39 3 538.44 25.480 4.73 432.40 544.70 565.00 132.600

F3 39 3 779.62 47.290 6.07 598.60 791.40 818.00 219.400

Individual Wt Mean mg

Circum Mean mm

PD (fully encapsulated)

Coresta Sub Group



Page 7 of 42

Filter Rod Individual Weights

Laboratory N did not measure weights on any filter samples.

Precision Statistics Individual weights data were screened for outliers and precision statistics calculated for each filter sample.

Table 3. Filter Rod Individual Weights Precision Statistics.

Individual value plots for each sample are shown in Fig.1. Individual data points are represented by solid symbols, and laboratory means by open circles.

Mandel Plots

The h statistic Mandel Plots (Fig.2) show that no laboratory is significantly different.

There are no apparent exceptional values in the k – statistic plots in Fig. 3.

Repeatability Variation

Cochran’s test did not identify any exceptional variances.

Z – scores (Fig.4)

Results for Laboratory C are questionable for all filters.

Filter Type Sample ID Mean sr sR r R CoV

SemiSlim F1 510.5 1.48 3.77 4.14 10.56 0.74

Slim F2 461.5 1.27 3.66 3.56 10.25 0.79

KS F3 1044.2 2.92 8.00 8.18 22.40 0.77

KS F4 798.9 3.36 6.49 9.41 18.17 0.81

KS F5 894.4 2.54 6.40 7.11 17.92 0.72

Coresta Sub Group



Page 8 of 42

Figure 1. Filter Rod Individual Weights – Individual Value Plots

Individual Values represented by solid symbols, mean values represented by open circles.

NJMIFEAHGPKDCBL

520

510

500

Indiv

idual W

t M

ean m

g

1

2

3

4

5

6

Inst_Code

NJMIFEAHGPKDCBL

465

460

455

Indiv

idual W

t M

ean m

g

1

2

3

4

5

6

Inst_Code

NJMIFEAHGPKDCBL

1050

1040

1030

Indiv

idual W

t M

ean m

g

1

2

3

4

5

6

Inst_Code

NJMIFEAHGPKDCBL

810

800

790

Indiv

idual W

t M

ean m

g1

2

3

4

5

6

Inst_Code

NJMIFEAHGPKDCBL

900

890

880

Indiv

idual W

t M

ean m

g

1

2

3

4

5

6

Inst_Code

Sample_ID = F1 Sample_ID = F2


Sample_ID = F5

Coresta Sub Group



Page 9 of 42

Figure 2. Filter Rod Individual Weights – h statistic Mandel Plots

Sample F5F4F3F2F1

2

1

0

-1

-2

h

kcrit

kcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab

Lab C ode PMLKJIHGFEDCBA

2

1

0

-1

-2

h

hcrit

hcritF 1

F2

F3

F4

F5

Sample

Individual Weight h statistics - Labs within Samples

Individual Weight h statistics - Samples within Labs

Coresta Sub Group



Page 10 of 42

Figure 3. Filter Rod Individual Weights – k statistic Mandel Plots

Sample F5F4F3F2F1

2.0

1.5

1.0

0.5

0.0

k

kcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab


2.0

1.5

1.0

0.5

0.0

k

kcritF 1

F2

F3

F4

F5

Sample

Individual Weight k statistics - Labs within Samples

Individual Weight k Statistics - Samples within Labs

Coresta Sub Group



Page 11 of 42

Figure 4. Filter Rod Individual Weights – z -Score Plots

CKBMFAJHEDLPGI

3

2

1

0

CBKMIDFPALHGJE

3

2

1

0

CBKMDAIFPJGLHE

3

2

1

0

CKBDIMFJEAHGLP

3

2

1

0

CKMBADPFGIHLEJ

3

2

1

0

F1 F2

F3 F4

F5

Coresta Sub Group



Page 12 of 42

Filter Circumference

Table 4. Filter Circumference Precision Statistics.

Results for Lab N excluded. One outlier removed sample F3, laboratory J.


Mandel Plots

Mandel plots do not show any laboratories beyond critical values (Note Laboratory N was not included in any of the filter analysis, and one measurement removed from sample F3, laboratory J).


Cochran’s test did not show any labs as outliers (one measurement removed form laboratory J).

Z – scores

Results for Laboratory C (Fig.8) are questionable for Filters F1 and F2.

Results for Laboratory B, Filters F2 and F4 are questionable.

Laboratories with missing data points did not, or were unable to, measure the samples.


SemiSlim F1 21.180 0.015 0.034 0.041 0.096 0.16

Slim F2 18.000 0.009 0.028 0.024 0.078 0.16

KS F3 24.440 0.016 0.037 0.044 0.104 0.15

KS F4 24.430 0.014 0.038 0.039 0.106 0.15

KS F5 24.470 0.019 0.037 0.052 0.102 0.15

Coresta Sub Group



Page 13 of 42

Figure 5. Filter Rod Circumference – Individual Value Plots

JMIFEAHGPKDCBL

21.20

21.15

21.10

Circum

Mean m

m

1

2

3

4

5

Inst_Code

JMIFEAHGPKDCBL

18.05

18.00

17.95

Circum

Mean m

m

1

2

3

4

5

Inst_Code

JMIFEAHGPKDCBL

24.45

24.40

24.35

Circum

Mean m

m

1

2

3

4

5

Inst_Code

JMIFEAHGPKDCBL

24.45

24.40

24.35

Circum

Mean m

m

1

2

3

4

5

Inst_Code

JMIFEAHGPKDCBL

24.56

24.48

24.40

Circum

Mean m

m

1

2

3

4

5

Inst_Code



Sample_ID = F5

Coresta Sub Group



Page 14 of 42

Figure 6. Filter Rod Circumference: h – statistic Mandel Plots

Sample F5F4F3F2F1

2

1

0

-1

-2

h

hcrit

hcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab


2

1

0

-1

-2

h

hcrit

hcritF 1

F2

F3

F4

F5

Sample

Filter Circumference h Statistics - Labs within Samples

Filter Circumference h Statistics - Samples within Labs

Coresta Sub Group



Page 15 of 42

Figure 7. Filter Rod Circumference: k – statistic Mandel Plots

Sample F5F4F3F2F1

2.0

1.5

1.0

0.5

0.0

k

kcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab


2.0

1.5

1.0

0.5

0.0

k

kcritF 1

F2

F3

F4

F5

Sample

Filter Circumference k - Statistics - Labs within samples

Filter Circumference k - Statistics - Samples within labs

Coresta Sub Group



Page 16 of 42

Figure 8. Filter Rod Circumference z -score Plots

CBAKDPMHEFLGIJ

2

1

0

Z_F1

BCDFAMKPLHJIGE

2

1

0

Z_F2

CBKHGMEADJPILF

2

1

0

Z_F3

BCKAPFMDGHILJE

2

1

0

Z_F4

CMAKGLBFPIHDJE

2

1

0

LAB ID_F5

Z_F5

F1 F2

F3 F4

F5

Coresta Sub Group



Page 17 of 42

FILTER ROD PRESSURE DROP

Table 5. Filter Rod Pressure Drop Precision Statistics

Laboratories N and B not included.


SemiSlim F1 400.030 2.603 6.770 7.288 18.956 1.69

Slim F2 416.170 2.297 7.720 6.432 21.616 1.86

KS F3 792.420 5.513 12.880 15.436 36.064 1.63

KS F4 244.620 2.519 3.960 7.053 11.088 1.62

KS F5 545.230 3.266 6.950 9.145 19.460 1.27


Mandel Plots

The h statistic Mandel Plots (Fig.10) show that Laboratory B is significantly different from the grand mean for all filters.

The k statistic plot (Fig.11) shows that Laboratory B’s within – laboratory variation is significantly different for all samples


Cochran’s test identifies the results for all filters for laboratory B as outliers.

Z – scores

Results for all filters for Laboratory B were identified as outliers


Coresta Sub Group



Page 18 of 42

Figure 9. Filter Rod Pressure Drop – Individual Value Plots

PMLKJIHGFEDCBA

400

375

350

PD

(fu

lly e

ncapsula

ted)

Mean m

m

J

K

L

M

P

A

B

C

D

E

F

G

H

I

Lab_Code

PMLKJIHGFEDCBA

420

400

380

PD

(fu

lly e

ncapsula

ted)

Mean m

m

J

K

L

M

P

A

B

C

D

E

F

G

H

I

Lab_Code

PMLKJIHGFEDCBA

800

700

600

PD

(fu

lly e

ncapsula

ted)

Mean m

m

J

K

L

M

P

A

B

C

D

E

F

G

H

I

Lab_Code

PMLKJIHGFEDCBA

250

240

230

PD

(fu

lly e

ncapsula

ted)

Mean m

m

J

K

L

M

P

A

B

C

D

E

F

G

H

I

Lab_Code

PMLKJIHGFEDCBA

550

500

450PD

(fu

lly e

ncapsula

ted)

Mean m

m

J

K

L

M

P

A

B

C

D

E

F

G

H

I

Lab_Code



Sample_ID = F5

Coresta Sub Group



Page 19 of 42

Figure 10. Filter Rod Pressure Drop: h - statistic Mandel Plots

F5F4F3F2F1

2

0

-2

-4

h

hcrit

hcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab

PMLKJIHGFEDCBA

2

0

-2

-4

h

hcrit

hcrit

F 1

F2

F3

F4

F5

Sample

Filter PD h Statistics - Labs within Samples

Filter PD h Statistics -Samples within Labs

Coresta Sub Group



Page 20 of 42

Figure 11. Filter Rod Pressure Drop: k - statistic Mandel Plots

F5F4F3F2F1

2.4

1.8

1.2

0.6

0.0

k

kcrit

J

K

L

M

P

A

B

C

D

E

F

G

H

I

C ode

Lab

PMLKJIHGFEDCBA

2.4

1.8

1.2

0.6

0.0

k

kcrit

F 1

F2

F3

F4

F5

Sample

Filter PD k Statistics - Labs within Samples

Filter PD k Statistics - Samples within Labs

Coresta Sub Group



Page 21 of 42

Filter Rod Pressure Drop – Repeatability Variation

Laboratory G did not submit data. All samples from Laboratory B were outliers.

Figure 12. Filter Rod Pressure Drop: z – score Plots

GBPAMICDHFLJEK

3

2

1

0

Z_Score

LGBICAMPDKHEFJ

3

2

1

0

Z_Score

GBAIMPDKFLJHEC

3

2

1

0

Z_Score

GBACILHMJEFPDK

3

2

1

0

Z_Score

GBAIMPEKFCDHJL

3

2

1

0

Z_Score

F1 F2

F3 F4

F5

Coresta Sub Group



Page 22 of 42

CIGARETTE DATA ANALYSIS

Laboratory P did not measure cigarettes. Laboratory N only measured circumference, but the values are not included in the analysis because they are wildly different from the rest of the laboratories. Laboratory I did not report PDo or Ventilation results.

Table 6. Cigarette Summary Statistics

Instrument Manufacturer/ Type Codes for Individual Value Plots

Instrument Manufacturer/Type Code

Borgwaldt 1

Cerulean 2

Sodiline 3

Sodimat 4

Sodimax 5

Variable Sample_ID N N* Mean StDev CoefVar Minimum Median Maximum

C1 39 0 954.86 13.410 1.4 921.00 954.00 993.90

C2 39 0 958.76 12.270 1.3 936.00 959.00 986.00

C3 39 0 716.26 10.620 1.5 688.00 717.00 741.60

C4 39 0 664.49 14.390 2.2 636.00 666.00 695.50

C5 39 0 657.69 12.580 1.9 633.00 661.00 680.70

C1 39 0 24.98 0.488 2.0 24.80 24.90 27.92

C2 39 0 25.01 0.491 2.0 24.84 24.94 27.97

C3 39 0 21.46 0.075 0.4 21.24 21.45 21.71

C4 39 0 18.02 0.027 0.2 17.97 18.01 18.08

C5 39 0 18.05 0.324 1.8 17.93 17.99 19.97

C1 36 3 118.71 6.200 5.2 100.10 119.75 127.80

C2 36 3 108.61 4.270 3.9 95.90 108.95 117.50

C3 36 3 120.00 3.130 2.6 115.00 119.50 126.80

C4 36 3 182.98 9.100 5.0 175.00 180.05 212.40

C5 36 3 98.32 31.250 31.8 85.00 88.75 204.10

C1 36 3 20.86 6.580 31.5 17.43 19.00 42.70

C2 36 3 29.53 5.780 19.6 26.53 27.84 48.70

C3 36 3 42.58 2.844 6.7 40.60 41.80 52.30

C4 36 3 46.10 4.570 9.9 41.85 45.05 61.20

C5 36 3 81.17 0.836 1.0 78.80 81.40 82.60

Individual Wt Mean mg

Circum Mean mm

PD open Mean mmWG

Vent % Mean

Coresta Sub Group



Page 23 of 42

CIGARETTE INDIVIDUAL WEIGHTS

Table 7. Cigarette Individual Weights – Precision Statistics

Individual value plots for each sample are shown in Fig.13. Individual data points are represented by open circles, and laboratory means by solid circles.

Mandel Plots

The h statistic Mandel Plots (Fig.14) do not show any laboratories significantly different from the reference value.

The k statistic plot (Fig.15) does not suggest exceptional variation for any single laboratory.


Cochran’s test shows repeatability variation is acceptable for all laboratories.

Z – scores

Results for Laboratory D (Fig.16), cigarette 1 and 3 are questionable. Results for Laboratory C are questionable.

There are no questionable results for cigarettes 2, 4 and 5.


Cigarette Type Sample ID Mean sr sR r R CoV

SuperSlim C5 657.7 4.53 12.58 12.68 35.22 1.9

SuperSlim C4 664.5 6.57 14.39 18.40 40.29 2.2

Slim C3 715.7 4.38 6.53 12.26 18.28 0.9

KS C1 955.5 7.09 9.86 19.85 27.61 1.0

KS C2 958.8 0.57 12.27 1.59 34.36 1.3

Coresta Sub Group



Page 24 of 42

Figure 13. Cigarette Individual Weights – Individual Value Plot

JMIFEAHGKDCBL

1000

960

920

Indiv

idual W

t M

ean m

g

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

980

960

940

Indiv

idual W

t M

ean m

g

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

740

720

700

Indiv

idual W

t M

ean m

g

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

700

675

650

Indiv

idual W

t M

ean m

g

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

680

660

640

Indiv

idual W

t M

ean m

g

1

2

3

4

5

InstCode

Sample_ID = C1 Sample_ID = C2


Sample_ID = C5

Coresta Sub Group



Page 25 of 42

Figure 14. Cigarette Individual Weights: h – statistic Mandel Plots

C 5C 4C 3C 2C 1

2

1

0

-1

-2

h

hcrit

hcrit

0

J

K

L

M

A

B

C

D

E

F

G

H

I

C ode

Lab

MLKJIHGFEDCBA

2

1

0

-1

-2

h

hcrit

hcrit

0

C 1

C 2

C 3

C 4

C 5

Sample

Individual Weights h statistics - Labs within Samples

Individual Weights h statistics - Samples within Labs

Coresta Sub Group



Page 26 of 42

Figure 15. Cigarette Individual Weights: k – statistic Mandel Plots

C 5C 4C 3C 2C 1

2.0

1.5

1.0

0.5

0.0

k

kcrit

J

K

L

M

A

B

C

D

E

F

G

H

I

C ode

Lab

MLKJIHGFEDCBA

2.0

1.5

1.0

0.5

0.0

k

kcritC 1

C 2

C 3

C 4

C 5

Sample

Individual Weight k statistics - Labs within Samples

Individual Weight k statistics- Samples within Labs

Coresta Sub Group



Page 27 of 42

Cigarette Individual Weights; Repeatability Variation

Cochran’s test did not show any laboratories with exceptional variation.

Figure 16. Cigarette Individual Weights: z -score Plots

DCBMKJLIEHFGA

3

2

1

0

LAB ID

Z S

core

Good

Satisfactory

Questionable

DBCKFMLGHEJIA

3

2

1

0

LAB ID

Z S

core

Questionable

Satisfactory

Good

DCBLEKMIJAGHF

3

2

1

0

LAB ID

Z S

core

Good

Satisfactory

Questionable

CDBFJMKHAGLEI

3

2

1

0

LAB ID

Z S

core

Questionable

Satisfactory

Good

DCKBMIJLFGHEA

3

2

1

0

LAB ID

Z S

core

Good

Satisfactory

Questionable

Sample C1 Sample C2

Sample C3 Sample C4

Sample C5

Coresta Sub Group



Page 28 of 42

CIGARETTE CIRCUMFERENCE

Table 8. Cigarette Circumference Precision Statistics

Individual value plots for each sample are shown in Fig. 17. Individual data points are represented by open circles, and laboratory means by solid circles.

Mandel Plots

The h statistic Mandel Plots (Fig.18) shows that Laboratory L is significantly different for cigarette C5

The k statistic plot (Fig.19) identifies Laboratory A as significantly different for cigarette C3, and Laboratory L for cigarette C5.


Cochran’s test shows Laboratory A cigarette C3, and Laboratory L, cigarette C5 as outliers, and Laboratory A cigarette C4 as straggler.

Z – scores

Results for Laboratory L, cigarette C5 (Fig.20) are highly questionable, results for Laboratory A cigarettes C1 and C2, are questionabel.



SuperSlim C5 17.99 0.020 0.026 0.057 0.072 0.14

SuperSlim C4 18.02 0.024 0.027 0.066 0.076 0.15

Slim C3 21.45 0.022 0.046 0.063 0.129 0.21

KS C1 24.90 0.037 0.058 0.104 0.162 0.23

KS C2 24.92 0.040 0.051 1.588 0.143 0.21

Coresta Sub Group



Page 29 of 42

Figure 17. Cigarette Circumference Individual Value Plots

JMIFEAHGKDCBL

25.0

24.9

24.8

Circum

Mean m

m

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

25.1

25.0

24.9

Circum

Mean m

m

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

21.52

21.44

21.36

Circum

Mean m

m

1

2

3

4

5

InstCode

InstCode

Lab_Code

54321

JMIFEAHGKDCBL

18.08

18.04

18.00

Circum

Mean m

m

1

2

3

4

5

InstCode

JMIFEAHGKDCBL

18.05

18.00

17.95

Circum

Mean m

m

1

2

3

4

5

InstCode



Sample_ID = C5

Coresta Sub Group



Page 30 of 42

Figure 18. Cigarette Circumference: h – statistics Mandel Plots

Sample C 5C 4C 3C 2C 1

4

2

0

-2

h

hcrit

hcrit

J

K

L

M

A

B

C

D

E

F

G

H

I

C ode

Lab

Lab C ode MLKJIHGFEDCBA

4

2

0

-2

h

-2.414

0

hcrit

C 1

C 2

C 3

C 4

C 5

Sample

Cigarette Circumference h Statistics - Labs within Samples

Cigarette Circumference h Statistics - Samples within Labs

Coresta Sub Group



Page 31 of 42

Figure 19. Cigarette Circumference: k – statistics Mandel Plots

Lab C ode MLKJIHGFEDCBA

3

2

1

0

k

kcrit

C 1

C 2

C 3

C 4

C 5

Sample

Sample C 5C 4C 3C 2C 1

3

2

1

0

k

kcrit

J

K

L

M

A

B

C

D

E

F

G

H

I

C ode

Lab

Cigarette Circumference k - Statistics - Samples within Labs

Cigarette Circumference k -Statistics - Labs within Samples

Coresta Sub Group



Page 32 of 42

Cigarette Circumference – Repeatability Variation

Laboratory A cigarette C3 and Laboratory L, cigarette C3 as outliers, and Laboratory A cigarette C4 as a straggler.

Figure 20. Cigarette Circumference: z -score Plots

ACLBGKIFDMHJE

3

2

1

0

Z S

core

ABCLGHIMKFDJE

3

2

1

0

Z S

core

CAHBLGKIFDMJE

3

2

1

0

Z S

core

GCIBHLKDEAMJF

3

2

1

0

Z S

core

LAKICFGEHJMDB

3

2

1

0

Z S

core

Plot of Z Score Cigarette C1 Plot of Z Score Cigarette C2

Plot of Z Score Cigarette C3 Plot of Z Score Cigarette C4

Plot of Z Score Cigarette C5

Coresta Sub Group



Page 33 of 42

CIGARETTE PRESSURE DROP (OPEN)

Table 9. Cigarette Pressure Drop (Open) – Precision Statistics

The Individual values plots for cigarette pressure drop identified Laboratory C as different enough to distort the scale on the graphs. The plots were redrawn without these values (Fig.21).

Mandel Plots

The h statistic Mandel Plots (Fig.22) shows that Laboratory C is significantly different from the reference value, for all cigarettes except C3.

The k statistic plot (Fig.23) does not suggest exceptional variation for any single laboratory.


Cochran’s test did not identify any outliers.

Z – scores

Laboratory C gave questionable results for cigarettes C1, C2, C4 and C5 (Fig.24). There were no questionable results for cigarette C3



SuperSlim C5 88.8 1.40 2.95 3.92 8.26 3.32

SuperSlim C4 180.6 3.35 4.35 9.38 12.18 2.41

Slim C3 120.0 1.57 3.13 4.40 8.76 2.61

KS C1 120.3 1.48 3.18 4.14 8.90 2.64

KS C2 109.4 1.19 2.23 3.33 6.24 2.04

Coresta Sub Group



Page 34 of 42

Figure 21. Cigarette Pressure Drop: Individual Value Plots

Labs N. I. P did not measure cig Pressure Drop, Lab C results removed from individual value plots.

JMIFEAHGKDBL

125

120

115

PD

open M

ean m

mW

G

1

2

3

4

5

InstCode

JMIFEAHGKDBL

115

110

105

PD

open M

ean m

mW

G

1

2

3

4

5

InstCode

JMIFEAHGKDBL

125

120

115

PD

open M

ean m

mW

G

1

2

3

4

5

InstCode

JMIFEAHGKDBL

192

184

176

PD

open M

ean m

mW

G

1

2

3

4

5

InstCode

JMIFEAHGKDBL

96

90

84

PD

open M

ean m

mW

G

1

2

3

4

5

InstCode



Sample_ID = C5

Coresta Sub Group



Page 35 of 42

Figure 22. Cigarette Pressure Drop: h – statistics Mandel Plots

C 5C 4C 3C 2C 1

4

2

0

-2

h

hcrit

hcrit

K

L

M

A

B

C

D

E

F

G

H

J

C ode

Lab

MLKJHGFEDCBA

4

2

0

-2

h

hcrit

hcrit

C 1

C 2

C 3

C 4

C 5

Sample

Pressure Drop (open) h statistics - Labs within Samples

Pressure Drop (open) h statistics - Samples within Labs

Coresta Sub Group



Page 36 of 42

Figure 23. Cigarette Pressure Drop: k – statistics Mandel Plots

MLKJHGFEDCBA

2.0

1.5

1.0

0.5

0.0

k

kcritC 1

C 2

C 3

C 4

C 5

Sample

C 5C 4C 3C 2C 1

2.0

1.5

1.0

0.5

0.0

k

kcrit

K

L

M

A

B

C

D

E

F

G

H

J

C ode

Lab

Pressure Drop (open) k statistics - Samples within Labs

Pressure Drop (open) k statistics - Labs within Samples

Coresta Sub Group



Page 37 of 42

Cigarette Pressure Drop (open) – Repeatability Variation: Cochran’s Test

Cochran’s test identified Labs B and C as stragglers for sample C5. There were no unusual results for the other 4 samples

Figure 24. Cigarette Pressure Drop: z - score Plots

CBAMEKLDJGFH

3

2

1

0

Z S

core

Good

Satisfactory

Questionable

CBAGMHLEFJKD

3

2

1

0

LAB ID_C2

Z S

core

Questionable

Satisfactory

Good

DLCBKAEJGHFM

3

2

1

0

LAB ID_C3

Z S

core

Good

Satisfactory

Questionable

CHADLEFJBKMG

3

2

1

0

LAB ID_C4

Z S

core

Questionable

Satisfactory

Good

CLHEAFJGKMBD

3

2

1

0

LAB ID_C5

Z S

core

Good

Satisfactory

Questionable

Cig C1 Cig C2

Cig C3 Cig C4

Cig C5

Coresta Sub Group



Page 38 of 42

CIGARETTE VENTILATION

Table 10. Cigarette Ventilation: Precision Statistics

The Individual values plots for cigarette pressure drop identified Laboratory C as different enough to distort the scale on the graphs. The plots were redrawn without these values (Fig.25).

Mandel Plots

Laboratory C was excluded from the Mandel plotsbecause of distortion of the graphs.

The h statistic Mandel Plots (Fig.26) shows that Laboratory B is significantly different from the reference value, for cigarettes C4 and C5. Laboratories A and B, in general, show a greater deviation form the reference values

The k statistic plot (Fig.27) does not shows any exceptional variation


Cochran’s test did not suggest any exceptional variation.

Z – scores

Results for Laboratory C (Fig.28) are highly questionable for C1, C2 , C3 and C4, and questionable for C5.

Results for Laboratory A , cigarette C2 are questionable.

Results for Laboratory B cigarette C4 are questionable.

Laboratories with missing data points did not, or were unable to, measure the samples, or had extreme values which distorted the plots.


SuperSlim C5 81.4 0.394 0.498 1.103 1.394 0.61

SuperSlim C4 44.9 0.643 0.938 1.800 2.626 2.09

Slim C3 41.8 0.334 0.428 0.935 1.198 1.02

KS C2 27.8 0.567 0.861 1.588 2.411 3.09

KS C1 18.9 0.399 0.573 1.117 1.604 3.03

Coresta Sub Group



Page 39 of 42

Figure 25. Cigarette Ventilation: Individual Value Plots

Laboratory C excluded from graphs because of wildly different results.

JMIFEAHGKDBL

20

19

18

Vent

% M

ean

1

2

3

4

5

InstCode

JMIFEAHGKDBL

30

28

26

Vent

% M

ean

1

2

3

4

5

InstCode

JMIFEAHGKDBL

42.4

41.6

40.8

Vent

% M

ean

1

2

3

4

5

InstCode

JMIFEAHGKDBL

46

44

42

Vent

% M

ean

1

2

3

4

5

InstCode

JMIFEAHGKDBL

82

81

80

Vent

% M

ean

1

2

3

4

5

InstCode



Sample_ID = C5

Coresta Sub Group



Page 40 of 42

Figure 26. Cigarette Ventilation: h – statistics Mandel Plots

C 5C 4C 3C 2C 1

3.0

1.5

0.0

-1.5

-3.0

h

hcrit

hcrit

K

L

M

A

B

C

D

E

F

G

H

J

C ode

Lab

MLKJHGFEDCBA

3.0

1.5

0.0

-1.5

-3.0

h

hcrit

hcrit

C 1

C 2

C 3

C 4

C 5

Sample

Ventilation h Statistic - Labs within Samples

Ventilation h Statistic - Samples within Labs

Coresta Sub Group



Page 41 of 42

Figure 27. Cigarette Ventilation: k – statistics Mandel Plots

C 5C 4C 3C 2C 1

2.0

1.5

1.0

0.5

0.0

k

kcrit

K

L

M

A

B

C

D

E

F

G

H

J

C ode

Lab

MLKJHGFEDCBA

2.0

1.5

1.0

0.5

0.0

k

kcritC 1

C 2

C 3

C 4

C 5

Sample

Ventilation k Statistic - Labs within Samples

Ventilation k Statistic - Samples within Labs

Coresta Sub Group



Page 42 of 42

Cigarette Ventilation: Repeatability Variation

All results were acceptable using Cochran’s test.

Figure 28. Cigarette Ventilation: z –score Plots

1 ISO/IEC Guide 43 – 1: 1997: Proficiency testing by interlaboratory comparisons – Part 1:

Development and operation of proficiency testing schemes

2 ISO/IEC 17025: 2005 General requirements for establishing the competence of calibration and

testing laboratories

3 ISO 5725 – 2 Accuracy (trueness and precision) of measurement methods and results — Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method

4 ibid.

CBAGDELFKJHM

3

2

1

0

LAB ID_C1

Z S

core

Good

Satisfactory

Questionable

Highly Questionable

CABHDJFMKELG

3

2

1

0

LAB ID_C2

Z S

core

Highly Questionable

Questionable

Satisfactory

Good

CBJLMAKDGFEH

3

2

1

0

LAB ID_C3

Z S

core

Good

Satisfactory

Questionable

Highly Questionable

CBAFMGJDKLEH

3

2

1

0

LAB ID_C4

Z S

core

Highly Questionable

Questionable

Satisfactory

Good

CBALDFHMKEGJ

3

2

1

0

LAB ID_C5

Z S

core

Good

Satisfactory

Questionable

Highly Questionable

Plot of C1 Z Scores Plot of C2 Z Scores

Plot of C3 Z Scores Plot of C4 Z Scores

Plot of C5 Z Scores

MMPTM Proficiency Test 2009(2) · Product Research Division, Japan Tobacco Inc., JT Sumida Building, 17-7, Yokogawa 1-chome, Sumida-ku, Tokyo 130-8603, Japan JT International Germany

Documents