58 



3. Analyses of Nicotine/Tar Ratios: 



The nicotine/tar ratio data provided by the FTC was 

 parallel to that for the nicotine and tar, consisting of two 

 additional vectors: 1) Ratio(i,j) = mean ratio of 

 nicotine/tar in cigarettes tested in year i, tar category j; 

 and 2) SDR(i,j) » std. dev. of that ratio in year i, tar 

 category j. In other words, the ratio is computed for each 

 tested cigarette, providing a mean and a std dev. of the 

 ratio. The mean ratio is not just computed as the ratio of 

 mean nicotine over mean tar. In this data, the mean and the 

 std. dev. were computed, according to Joe Nulholland by 

 taking sales weighted means and sd's of all of the varieties 

 tested within a given year and tar category. 



We did five analyses of the nicotine/tar ratio. These were 

 as follows: 



1) weighted least squares regression of the ratio, 

 Ratio(i,j) on year i, stratified by tar category j - (high, 

 low, and ultra-low) . The weights were proportional to 

 reciprocals of the variances of the sample means > 

 l/VarR(i,j) - N(i,j)/SDR(i,j)-2. 



2) ordinary (unweighted) least squares regression of the 

 ratio on year, stratified by tar category. (This was the 

 same as analysis 1 except the weights were now ■ 1/10.) 



3) Spearman rank correlation of the ratio and year, 

 stratified by tar category. 



4) Analysis of Covariance, using tar category as the class 

 variable and year as a continuous covariate. Tukey tests 

 were run at simultaneous level .05 to determine which tar 

 categories had significantly different nicotine/tar ratios 

 when adjusted for year. This analysis treated the 30 observed 

 Ratio(i,j) as single points with equal weights. 



5) Analysis of Covariance, using tar category as the class 

 variable and year as a continuous covariate. Again, Tukey 

 tests were run at simultaneous level .05 to determine which 

 tar categories had significantly different nicotine/tar 

 ratios when adjusted for year. This analysis gave the 30 

 observed Ratio(i,j) weights proportional to the reciprocals 



