*• > 



490 



-♦- 



la which tMpoaaeot* v«r« ,»*kcd Co rAok cb«lr ujual VrAoa tod Mch of 

 ch« tvo tctt cls^cct** iB cb«lr order of prtf«r«nc«. 

 KOUHP ROBIM AMALWl 



Round Robtfr-Analyrla it a aucliclcal tccfani^uc which iofcrt a 

 relation«hip becv^^^ in cbia cla«, cvo cigarcccca of different olcotiM 

 delivery which were noc^^etted againac each ocher. la aiople ccraa, 

 it ia assuffled Chat if a smojter preferred A to B and preferred B to C 

 then the net preference for A>9«r B plua the net preference for 

 B over C will equal the net preference for k over C to within the li«ita 

 eMected from the variability of the todividoal aeasureoeDts. The 

 result is that the effective sample size La increated provided the 



V "'' 



acatistic4l test for additivity iodicatea that this ia a valid calculation. 



^? ^/ 



Ia this ^rvcy it was impossible to ask •■vtif respondent to try 

 ■ ^'y 'C 



all cofflblaacions^'Of the four nicotine levels against c^^ oCb«r, so 

 Che Round Robin was 4«jtd to naxiaise the data use. The aOicivlc/ test 

 indicated that this was Cvalid test using the current data. 



Round Robia Aralyala wx1r;applied only to Che prefereoce resulcs 

 and not to any of the attribute '^Mles (i.e. strength, smoothness, etc.). 

 The variables used arc shown la Appendix V. 



ipe 

 RESULTS AND DISCUSSION V 



r 



I. FIRST TEST - PRODUCTS A. B. C, 



Regular Brjnd Data '7^^ 



In order to have up to date information on thoit who actually smoked 

 the test products, some of the questions relating to regdlar brand usage ^ 



that had already been asked 4 months previously during the "tiain survey •*•* 



•V - o 



were repeated. It was found that 91Z still smoVcd chc saata regular ^ 



o 



