Tobacco Documents Online

~Z -5- assessment parameters, environmental conditions with each other and assessment parameters with each other for both low delivery cigarettes (Appendix II). Also the transformed data within these tables were represented by a two-dimensional diagram, termed a DELTA plot (Appendix II)o This approach gave a graphical best fit summary of the general trends existing within the data matrix, Unfortunately from these results it is impossible to disentangle the complicated inter-relationships among the variables or to interpret trends within the DELTA plot. For this reason it was decided to consider an alternative multivariate approach and/or analysis of variance and covariance in an effort to find a more suitable method for data analysis and the determination of any effects due to changes in environmental conditions. Statistical Methodology The experiment consisted of five factors. (A) Temperature at five levels O, 10, 20, 30, 40°C. (B) Relative humidity at five levels, 15, 35, 55, 75, 95%. (C) Subject at nine levels. (D) Brand at two levels. (E) Session at 225 levels (temperature, relative humidity, subjects). The first four factors (A, B, C, D) are crossed with each other and will be considered as fixed effects. That is to O 1983 British American Tobacco Co. Ltd. This report must not be copied or shown to unauthorised persons. Co 00 Ob tO

i¸ :" :T • i~~• ~, .. I ! -6- say, only the levels of the factors actually in this experiment were considered. For example, the subjects are not considered to be a random sample of subjects who might have done the experiment. 1. Linear. Session is nested within •temperature, relative humidity and subject, but is crossed with brand. It is considered as a random effect, i.e. the one session for each subject under each environmental condition is considered to be a random sample of the many sessions that might have been. The two factors, temperature and relative humidity both have five equally spaced levels, and thus the effects of each are partitioned into three orthogonal components. r' i! 2. Quadratic. 3. Remainder or deviation. The first stage of the analysis was to examine the DELTA scores from the cigarettes smoked at ambient environmental conditions. An analysis of variance was performed on each assessment parameter and brand separately using the factors temperature, relative humidity and subject. This was done to establish that the ambient environmental condition scores were not affected by the forthcoming environmental chamber conditions. There were no effects of temperature and relative humidity on any of the assessment parameters and thus it was concluded that the subjects were not affected by anticipated future conditions. As a result of this the ambient environmental condition scores can be considered as potential covariates in the analysis of the environmental chamber scores. As there were nine assessment parameters it might be thought that a multivariate approach would be fruitful, especially if O 1983 British American Tobacco Co. Lid. This rcpo~ mus~ not be copied or shown to unauthoriscd persons.

/~ i~¸ .k -7- the assessment parameters are highly correlated with each other. The within subject correlation matrix for %he ambient environmental condition scores was computed for both brands separately and as these were very similar a pooled correlation matrix was calculated (Appendix Ill). All correlations were less than 0.4 in absolute value except -0.51 between Mouthful] and Draw Effort. These correlations are very similar to those reported in a previous GR&DC study (3). Hence as nearly all assessment parameters were at most weakly correlated a multivariate approach did not seem worthwhile. The next stage was an analysis of covariance for each of the nine assessment parameters scored in the environmental chamber with the ambient environmental condition scores as covariates. Initially this was done separately for both brands, which allowed judgement on whether or not it was valid to combine the brands. For this to be so the relationship between the assessment parameters scored in the environmental chamber with the assessment parameters scored at ambient environmental conditions should be similar for both brands, this was found to be the case for all assessment parameters. In addition the residual mean squares for both brands should be similar, which they were. Hence the two brands were combined. The final stage was to consider both brands in an analysis of covariance for each of the nine assessment parameters scored in the environmental chamber with the ambient environmental conditions scores as covariates (Appendix IV). There were two levels of analysis between and within sessions. The within level was solely concerned with the differences between the two brands and whether these differences were affected by the subject, temperature and relative humidity. The latter two are of considerable importance since if the effects of temperature and relative humidity were different for the two brands it would be imprudent to average the © 1983 British American Tobacco Co. Ltd, This report rnusZ not be copied or shown to unauthoriscd persons. C=T[:~

/ -8- effects over brands. However it was found that the effects of temperature and relative humidity were very similar for both brands. Hence it was desirable to average the effects over brands. Having found the effects of temperature and relative humidity were not dependent on brand, the between session level was considered. Initially the pair-wise interactions between temperature, relative humidity and subjects were examined. If the interaction between subject and temperature was large then the effect of temperature varied between the nine subjects, likewise relative humidity. This would give rise for concern since if another group of subjects had been used then quite different results might have occurred. Fortunately the interactions with subjects were not large. The interaction between temperature and relative humidity was also very important, since if it was large then the effects of temperature depend upon the level of relative humidity and vice versa. However if there was no interaction, then the effects of temperature and relative humidity were additive and the main effects could be considered. ! For each assessment parameter there were thirty-five significance tests involving either temperature or relative humidity. All of these were independent of one another, and thus even if there were no effects of temperature and relative humidity there would be some significant results. This is similar to the multiple comparisons problem when comparing several brands. The approach taken was informal. Main effects were given more credibility than two factor interactions, which were in turn preferred to three factor interactions. Similarly in the partitions for temperature and relative humidity the linear component was favoured over the quadratic component which had a priority over the deviation component. All analyses were done using GENSTAT (4). Unfortunately one of the subjects was unable to complete one session, and no qB L983 Brltish American Tobacco Co. Ltd. This report must not be. copied or shown to unauthorised persons. ,I • ~!.! Igg~,

L, : • ?7 . ,~E~ l -g- scores were obtained in the environmental chamber. The missing scores were estimated using the GENSTAT missing value procedure described in (5). RESULTS For each assessment parameter, the analysis of covariance is shown in Appendix III and the panel means are shown in Tables 2, 3, 4, 5, 8, 11, 12, 13, 14. In general the effects of temperature and relative humidity could not be distinguished between the two low delivery cigarettes and the nine subjects. The effects relating to each DELTA assessment parameter will be considered in turn. Mouthfull The effects of temperature on Mouthfull did not depend upon the level of relative humidity and vice versa. From Table 2 it can be seen that increasing temperature decreases Mouthful] at a rate of 0.0075 units per °C and there is no indication that the effect is not linear. Increasing relative humidity increases Mouthfull at a rate of 0.0028 units per % RH. However there is some indication of non-linearity as the average Mouthfull at 55% RH is less than might be expected. ~D 1983 British American Tobacco Co. L~d. This report must not be copied or shown to unauthorLsed persons. ca °

i = -10- TABLE 2 MOUTHFUL TABLES OF NEANS AFTER ADJUSTING FOR MOUTHFULL SCORES UNDER AFI~IENT CONDITIONS Temperature °C 0 10 ZO 30 40 SED* 3.82 3.77 3.70 3.60 3.52 0.09 Estimated linear effect of temperature is -0.0075 (0.0021) per degree C % Relative Humidity 15 35 55 75 95 SED 3.61 3.70 3.51 3.73 3.87 0.09 Estimated linear effect of relative humidity is 0.0028 (0.0010) per % RH Temperature °C 0 10 20 30 40 O.S. Blended 4.04 4.11 4.02 3.93 3.90 SED U.K. Virginia 3.59 3.43 3.39 3.28 3.15 0.13 % Relative Humidity 15 35 55 75 95 U.S. Blended 3.99 4.05 3.79 3.96 4.21 SED U.K. Virginil~ 3.22 3.35 3.23 3.50 3.53 0.13 % Relative Humidity 15 35 55 75 95 Temperature °C 0 3.5 3.9 3.7 3.9 4.1 I0 3.7 3.9 3.6 4.0 3.7 SED 20 3.9 3.6 3.4 3.7 4.0 0.21 30 3.5 3.6 3.9 3.3 3.7 - 40 3.4 3.5 3.0 3.8 3.8 Relative Humidity 15 35 55 75 95 Temperature °C U.S. Blended 0 3.8 4.3 3.8 4.0 4.5 10 4.3 4.3 4.0 4.1 3.9 20 4.1 3.9 3.6 4.0 4.5 30 3.7 3.9 4.1 3.7 4.2 40 4.0 3.8 3.5 4.1 4.0 SED U.K. Virginia 0 3.3 3.6 3.6 3.8 3.7 0.27 I0 3.2 3.5 3.1 3.9 3.4 20 3.6 3.2 3.2 3.4 3.6 30 3.3 3.2 3.7 2.9 3.3 40 2.8 3.3 2.5 3.5 3.6 * SED is the standard error of differences between any pair of means of the same brand. [983 British American Tobacco Co, Ltd. T~s report must not I~ copied or shown to unouthorlsed persons. .E • ~. UI m

• T~, ,H7 r~, 7. -11- Draw Effort From Table 3 it can be seen that for Draw Effort, the relationships with temperature and relative humidity are predominantly linear. When the level of temperature is increasing, Draw Effort increases at the rate of 0.0057 units per °C. Draw Effort decreases at the rate of 0.0026 units per % as relative humidity increases. i - !_ • ., .a . - ;J~ 1 a,': C ~ . © 1983 British American Tobacco Co. Lid. This report must not be copied or shown to unauthorised persons. ¢J 00

i. -12- TABLE 3 DRAW EFFORT TABLES OF ~ANS AFTER ADJUSTING FOR DRAW EFFORT SCORES UNDER ANBIENT CONDITIONS Temperature °C 0 I0 20 30 40 SED 2.05 2.23 2.23 2.26 2.32 0.09 Estimated linear effect of temperature is 0.0057 (0.0020) per degree C % Relative Humidity 15 35 55 75 95 SED 2.31 2.27 2.21 2.20 2.09 0.09 Estimated linear effect of relative humidity is -0.0026 (0.0010) per % RH Temperature °C 0 10 20 30 40 U.S. Blended 1.84 1.91 1.99 2.01 2.02 SED U.K. Virginia 2.26 2.55 2.47 2.50 2.61 0.13 % Relative Humidity 15 35 55 75 95 U.S. Blended 2.03 1.91 2.00 2.01 1.80 SED U.K. Virgin~( 2.59 2.63 2.42 2.3B 2.37 0.13 % Relative Humidity 15 35 55 75 95 Temperature °C 0 2.2 2.0 2.0 2.0 2.0 10 2.3 2.2 2.3 2.0 2.3 " 20 2.3 2.4 2.2 2.5 1.9 30 2.3 2.4 2.3 2.2 2.0 40 2.5 2.4 2.2 2.2 2.3 % Relative Humidity 15 35 55 75 95 Temperature °C U.S. Blended 0 1.9 1.7 1.9 2.0 1.6 10 2.0 1.9 2.0 1.6 2.0 • 20 2.0 2.1 2.0 2.3 1.6 30 2.1 1.9 2.3 2.1 1.7 40 2.1 2.0 1.8 2.1 2.1 U.K. Virginia 0 2.4 2.3 2.2 2.0 2.3 10 2.5 2.6 2.6 2.4 2.5 20 2.5 2.7 2.4 2.7 2.1 30 2.5 2.9 2.4 2.4 2.3 40 3.0 2.7 2.5 2.3 2.5 SED O. 20 SED 0.29 ID i983 British American Tobacco Co. Lid, This report mus', not b¢ copicd or shown to unau',horlsed persons. !j: •>% :m cA Go ob cn °

,i .~ ~ ~i-_"~ ~ :k -13- Mouthfeel Mouthfeel is linearly related to relative humidity, the estimated effect is -0.0034 units per % relative humidity (Table 4). This decrease in Mouthfeel as relative humidity increases did not depend upon temperature. Mouthfeel may be related to temperature, with the extremes O°C and 40°C producing a lower response than the intermediate temperatures. 1 © 1983 British American Tobacco Co. Ltd. This report must not be copied or shown to unauthor/scd persons. Go

-14- TABLE 4 MOUTHFEEL TABLES OF MEANS AFTER ADJUSTING FOR MOUTHFEEL SCORES UNDER Ar~BIENT CONDITIONS Temperature °C 0 10 20 30 40 SED 2.32 2.55 2.47 2.47 2.29 0.13 Estimated linear effect of temperature is -0.0014 (0.0030) per degree C % Relative Humidity 15 35 55 75 95 SED 2.55 2.52 2.40 2.33 2.30 0.13 Estimated linear effect of relative humidity is -0.0034 (0.0015) per % RH Temperature °C 0 10 20 30 40 U.S. Blended 2.52 2.84 2.59 2.81 2.68 SED U.K. Virginla 2.12 2.25 2.35 2.13 1.89 0.19 % Relative Humidity 15 35 55 75 95 U.S. Blended 2.81 2.78 2.62 2.58 2.65 SED U.K. Virginla 2.29 2.25 2.18 2.08 1.94 0.19 % Relative Humidity 15 35 55 75 95 Temperature °C 0 2.2 2.6 2.4 2.1 2.3 10 2.6 2.6 2.8 2.7 2.1 20 2.7 2.7 2.3 2.3 2.3 30 2.5 2.4 2.4 2.3 2.8 40 2.7 2.3 2.2 2.3 1.9 % Relative Humidity 15 35 55 75 95 Temperature °C U.S. Blended 0 2.5 2.5 2.5 2.4 2.7 10 2.9 3.0 2.9 2.9 2.5 20 3.0 3.0 2.2 2.2 2.5 30 2.7 2.6 2.6 2.7 3.4 40 2.9 2.7 2.9 2.7 2.2 U.K. Virginia 0 1.9 2.7 2.2 1.8 2,0 10 2.3 2.2 2.6 2.4 1.8 20 2.5 2.4 2.3 2.4 2.1 30 2.2 2.1 2.2 1.9 2.3 40 2.5 1.9 1.5 1.9 1.6 SED •0.30 SED 0.41 © 1983 Brilish American Tobacco Co. Lld, ThZs report must not be copied or shown to unauthorised persons.