Health Sciences 3801A/B Lecture Notes - Lecture 9: Observational Error, Eating Disorder, Repeated Measures Design
Document Summary
The more likely it is that the f-ratio is statistically significant, the greater the power of our analysis of variance. Trying to demonstrate a significant effect of the independent variable on the dependent variable - demonstrated when the treatment effect is significantly larger than the error effect. Increasing the magnitude of the treatment effect isn"t practical (says to measure larger differences) Decreasing the magnitude of the error effect, this allows us to achieve greater power. As the error goes down, the f-ratio goes up. If you"re looking at a dependent variable, there is substantial evidence to suggest that sex is a significant predictor of a drive for thinness (eating disorder) If you don"t match for sex - all of the error from sex will go into the unsystematic error term. It will be considered random error within the data. When you match variables, you can pull out the effects of sex (systematic error)