A trend statistic and its variance were computed for each contingency table Trends were added up and variances were added up separately for the fatal tumor type and for the incidental one.The asymptotic probability that a trend was greater than or equal to the observed dose related trend, was computed separately for each context of observation.Afterwards, the fatal and incidental trends were summed, and their variances were summed too; and the overall asymptotic one-tailed probability for that tumor type was computed.Mathematical formula's had been given by Peto(ref. 1; pp. 386 - 388).Following Peto's advice(ref. 1, p. 402)the continuity correction was not applied.Not making this correction resulted in lowering the p-value and in increasing the risk of a false positive.When the asymptotic p-value was at least marginally statistically significant(p 0.10)and when the total number of tumor bearing animals in all treatment groups was 8 or less(ref. 3 p. 146), the "exact" age-adjusted Cochran-Armitage trend was computed also(ref. 4)and it was preferred.The resulting p-value was the "exact" probability that a trend was greater than or equal to the one observed.For computer technical reasons, experimental periods in tables are expressed in lunar months(28 days or 4 weeks).Computer print-outs refer mostly to a carcinogenicity study of 24 calendar months duration as originally planned.The study was extended to 25 calendar months or approximately 28 lunar months(111 weeks).- report archive(Building 070)- specimen archive(Building 162)- technical maintenance unit(Building 260).Test for positive dose related trend- Cumulative mortality rate per lunar monthAfter 24 calendar months(26 lunar months), mortality in the control group was 42% in males and 40% in females respectively.Termination of the study was decided after approximately 28 lunar months or 109-111 weeks(=25 calendar months)when mortality was about 50% in the control group and the low dose group of both males and females.When mortality rate(dead or sacrificed animals)in the dosed groups was compared with that of the control group(see pages R 1.3 - R 1.5), the following was noted:Trend analysis did not reveal a positive dose related trend in mortality rate Intergroup comparison, using the Chi-square test, showed a significantly(p