Final report : assessment of the impact of highway runoff on freshwater mussels in North Carolina streams - Page 32 |
Previous | 32 of 109 | Next |
|
small (250x250 max)
medium (500x500 max)
Large
Extra Large
large ( > 500x500)
Full Resolution
All (PDF)
|
This page
All
|
32 The glycogen standard curves for the tissue samples were linear in the test range described ( 10- 250 mg/ dL) with R- squared ( coefficient of determination) values of 99.7% and 99.6% respectively. The average % difference of duplicate samples was 12.23% ( range, 1.27 – 24.36). The mean RSD ( relative standard deviation) of triplicate in- house standards was 8.13% ( range, 5.44 – 11.61). Recovery of known additions averaged 91.08% ( range, 62.38 – 109.32). Statistical Analysis: Reference intervals and associated confidence limits for hemolymph parameters were calculated from the group of 380 animals using nonparametric determination of the central 95% intervals ( Solberg, 1986). The nonparametric procedures, based on ranks, alleviated analytical problems associated with data sets bounded by lower limits of detection and allowed out- of- bound data to retain full significance. The results for each parameter were sorted in ascending order and assigned a rank number from 1 to n. The bounds of the central 95% reference interval were then defined as the values corresponding to rank numbers equal to 0.025( n+ 1) and 0.975( n+ 1). If the calculated rank numbers were not integers, then values were interpolated from the two closest ranks. The 90% confidence intervals were calculated from the binomial distribution ( Solberg, 1986). For datasets of n = 373- 380, the 90% confidence interval for the lower 2.5 percentile corresponded to rank numbers 5 and 16. The 90% CI for the upper 97.5 percentile corresponded to rank numbers ( n+ 1)- 16 and ( n+ 1)- 5. For datasets of n = 364- 372, the 90% confidence interval for the lower 2.5 percentile corresponded to rank numbers 5 and 15, and the 90% CI for the upper 97.5 percentile corresponded to rank numbers ( n+ 1)- 15 and ( n+ 1)- 5. Out- of- bound data were not an issue for glycogen results, and presumed normality was not rejected by the Kolmogorov- Smirnov test at a 0.05 significance level ( p > 0.15). Consequently, parametric reference ranges, providing narrower distributions, are reported for glycogen. After ruling out conditional dependence between land- use designation and gravidity status or shell length, we used Spearman’s Rank correlation coefficient on the full data set of 380 animals to identify correlations between animal length and potential health parameters. We evaluated associations between potential health parameters and gravidity status logistic regression. Associations between site abundance/ diversity rank and site prevalence of unusual health measures were established using Chi- square tests and logistic regression. Bayesian likelihood ratios are presented as measures of strength of association ( Rothman and Greenland, 1998) between variables that may hold predictive value for future diagnosis of compromised unionid populations. A Bayesian likelihood ratio represents the prevalence of a characteristic among a compromised ( or test) group relative to the prevalence of the same characteristic among the healthy ( or control) group ( Gustafson et al., 1998). Effects of road crossings and landuse We used the nonparametric Kruskal- Wallis test to evaluate differences in health parameter medians between landuse types ( n = 10 forested and 9 agricultural sites), and between upstream ( above bridge) and downstream ( below bridge) collection locations ( n = 19 sites).
Object Description
Description
Title | Final report : assessment of the impact of highway runoff on freshwater mussels in North Carolina streams - Page 32 |
Pres Local File Path-M | \Preservation_content\StatePubs\pubs_borndigital\images_master\ |
Full Text | 32 The glycogen standard curves for the tissue samples were linear in the test range described ( 10- 250 mg/ dL) with R- squared ( coefficient of determination) values of 99.7% and 99.6% respectively. The average % difference of duplicate samples was 12.23% ( range, 1.27 – 24.36). The mean RSD ( relative standard deviation) of triplicate in- house standards was 8.13% ( range, 5.44 – 11.61). Recovery of known additions averaged 91.08% ( range, 62.38 – 109.32). Statistical Analysis: Reference intervals and associated confidence limits for hemolymph parameters were calculated from the group of 380 animals using nonparametric determination of the central 95% intervals ( Solberg, 1986). The nonparametric procedures, based on ranks, alleviated analytical problems associated with data sets bounded by lower limits of detection and allowed out- of- bound data to retain full significance. The results for each parameter were sorted in ascending order and assigned a rank number from 1 to n. The bounds of the central 95% reference interval were then defined as the values corresponding to rank numbers equal to 0.025( n+ 1) and 0.975( n+ 1). If the calculated rank numbers were not integers, then values were interpolated from the two closest ranks. The 90% confidence intervals were calculated from the binomial distribution ( Solberg, 1986). For datasets of n = 373- 380, the 90% confidence interval for the lower 2.5 percentile corresponded to rank numbers 5 and 16. The 90% CI for the upper 97.5 percentile corresponded to rank numbers ( n+ 1)- 16 and ( n+ 1)- 5. For datasets of n = 364- 372, the 90% confidence interval for the lower 2.5 percentile corresponded to rank numbers 5 and 15, and the 90% CI for the upper 97.5 percentile corresponded to rank numbers ( n+ 1)- 15 and ( n+ 1)- 5. Out- of- bound data were not an issue for glycogen results, and presumed normality was not rejected by the Kolmogorov- Smirnov test at a 0.05 significance level ( p > 0.15). Consequently, parametric reference ranges, providing narrower distributions, are reported for glycogen. After ruling out conditional dependence between land- use designation and gravidity status or shell length, we used Spearman’s Rank correlation coefficient on the full data set of 380 animals to identify correlations between animal length and potential health parameters. We evaluated associations between potential health parameters and gravidity status logistic regression. Associations between site abundance/ diversity rank and site prevalence of unusual health measures were established using Chi- square tests and logistic regression. Bayesian likelihood ratios are presented as measures of strength of association ( Rothman and Greenland, 1998) between variables that may hold predictive value for future diagnosis of compromised unionid populations. A Bayesian likelihood ratio represents the prevalence of a characteristic among a compromised ( or test) group relative to the prevalence of the same characteristic among the healthy ( or control) group ( Gustafson et al., 1998). Effects of road crossings and landuse We used the nonparametric Kruskal- Wallis test to evaluate differences in health parameter medians between landuse types ( n = 10 forested and 9 agricultural sites), and between upstream ( above bridge) and downstream ( below bridge) collection locations ( n = 19 sites). |