ENV330H5 Lecture Notes - Lecture 4: Heracleum Maximum
Document Summary
Random variation: many things influence fish length (genetics; early development, food, etc. ) If we can assume that random variation will be similar in both streams (i. e. uncorrelated with the independent variable), this random variation will introduce. Precision will be improved by taking many replicates. Different early life, genetic composition nothing to do with contamination just random variation. Bias: sampling technique may only reach part of the population of interest (e. g. nets only catch fish above a certain size) If the bias is similar in both streams it will introduce inaccuracy into the data, but the any differences between groups may still be detected. More replicates won"t solve the problem; need to think about the study design and questions asked to determine the severity of the error. May need to adjust inference to a sub-population rather than the whole: confounding/collinear variable: something else that affects fish length that varies with the independent variable.
