Aggregation of estimates 



The data for trips, catch, and participants were all calculated at the State, mode, area, 

 and species level. Estimates were made on a wave-by-wave basis with all wave estimates 

 calculated using the same formulas. The 1986 survey estimates were calculated by summing 

 six waves of data from January-December for Georgia, the east coast of Florida and the Gulf 

 coast States, and five waves of data from March-December for the Atlantic coast States north 

 of Georgia. Elimination of sampling during January and February on the Atlantic coast results 

 in an underestimate of less than 10 percent of the catch and number of trips for the calendar 

 year. The data used to produce the tables in this publication are maintained in their 

 unaggregated form in the MRFSS data base. 



SAMPLING VARIANCES 



A clustered sampling design was used throughout the telephone interview portion of the 

 survey. However, the variance associated with the average number of fishing trips was 

 calculated using a stratified simple random sample model for each stratum. 



Estimation of the variances associated with the average catch and weight estimates 

 obtained from the intercept survey was based on the assumptions that the primary sampling 

 unit was the fishing trip of a single person and that there was no clustering effect due to 

 grouping interviews at a site. These assumptions had been empirically verified in pilot 

 surveys. Therefore, the variance was estimated using the standard variance equation for a 



2 



stratified random sample: o-_ = £ 



where: 



and 



X. 



I 



2 



n. 

 I 



s = estimated variance of stratum i, 

 "j = number of interviews in stratum i. 



Estimation of the variance of the combined estimates for the two surveys required 

 special attention. The intercept survey provided estimates (xj) for the average catch per trip 

 for a mode within a State. The sampling variance v(xj) was estimated for the sample as 

 described above. Let Wj be the estimate of the total number of trips from the telephone zone 

 in the State for the corresponding mode (obtained from the telephone survey), and Pj be the 

 proportion of trips from the telephone zone to the total trips intercepted. Then yj - Wj/pj is 

 the estimate for the total number of trips within a State for a mode. Since y is a ratio of two 



