Conference on Applied Statistics in Agriculture

VALIDITY OF 95% t-CONFIDENCE INTERVALS UNDER SOME TRANSECT SAMPLING STRATEGIES

Soil pH data were used to assess the capture rates of 95 % t-confidence intervals based on five different transect sampling strategies. Two different sampling methods were considered, "deterministic" and "two-stage simple random sampling". The data used were pH readings at 15 and 30 centimeter depths from two local agricultural fields in the Manhattan, Kansas area. The data provided three distinct populations with three different distributions - skewed left, symmetric, and bimodal. The total number of transects randomly sampled was 2, 5, and 10. The total number of points sampled along each transect was 2, 7 and 14. The 95% t-confidence intervals were simulated 5000 times using five different transect sampling strategies, and the capture rates of the population mean were recorded. Box plots of the capture rates for the five transect sampling strategies were constructed and compared. In most cases the "deterministic" sampling method had capture rates that underestimated the 95 % confidence level, whereas the two-stage sampling strategies produced capture rates which were conservative or closer to the 95 % confidence level. The variances of the capture rates for the two-stage sampling strategies were relatively small in comparison to the deterministic sampling strategies. In conclusion, the "two-stage simple random sampling" method along with the Satterthwaite degrees of freedom are recommended when using transect sampling.