Design and Analysis of Salmonid Tagging Studies in the Columbia Basin, Volume IX; Comparison of Statistical Methods of Estimating Treatment-Control Ratios Using Coded-Wire Tags, Based on Spring Chinook Salmon on the Columbia River, 1986-1988 Technical Report.
The strength of a salmon run is often measured as the adult return rate from some previous brood year (i.e. the percent of a smolt population returning to spawn or captured in fisheries). The U.S. Army Corp of Engineers (ACOE) program of barge transportation of smolts from collector dams is one mitigation measure used to improve smolt survival. Using Coded Wire-tags, the adult return rates of transported and untransported smolt have been tracked. A ratio of the recovered percentages of adult salmon, those transported in the smolt stage over the salmon not transported (controls), is often used to summarize the program effectiveness. There are a number of ways to estimate this transportation/control (T/C) ratio, and this paper explores six alternative statistical models to improve accuracy and precision of the estimate. Assuming the proportion of adult recoveries are binomially distributed, the data were analyzed using linear regression of arc-sine square-root and logit transformations; general linear model regression (GLM) with logit- and log-links; and a maximum-likelihood estimation (MLE) of the T/C ratio. Profile likelihood intervals were calculated to generate 95% confidence interval estimates of the T/C ratio. Depending on the analytical method, T/C ratios varied greatly. Arc-sine square-root and logit transformations gave individual release T/C ratios which ranged from 1.0934 to 4.0076 and {minus}1.2193 to 1.9057, respectively. The negative T/C ratio is due to the back-transformation properties of the logit transformation. The GLM and MLE approaches produced mean T/C ratios (after adjusting for the individual release batch effects) ranging from 1.4964 to 1.4974. The recommended method from this analysis, a binomial maximum likelihood estimate adjusted for over-dispersion, produced a T/C ratio of 1.4965 with a 95% confidence interval of (1.0618, 1.9312).