We couldn't possibly cover all of the first semester's content in two days of review, so let's point out a few topics that no one has asked about:
Correlation coefficient r and coefficient of determination r-squared. What special insight does the value of r-squared give you about the relationship between x and y?
Why can't we just take the square root of r-squared to get the value of r?
Slope of a LSRL = the estimated increase (or decrease) in the response variable for every unit increase in the explanatory variable.
The easiest way to get it is r*sy/sx where sy is the sample standard deviation of y and sx is the sample standard deviation of x.
While we're talking about sample standard deviations. . . the formula is SQRT(variance of the variable), so the sample standard deviation of x would be
SQRT[(sum of the squares of (Xi - Xbar) for all values of X)/(n-1)]. N is the sample size.
Don't panic if you can't read that--just look up the formula in the text.
What does it mean to be resistant to outliers? Give examples of measures which are resistant. Give examples of some which are not.
What are the benefits of different types of graphs (box and whisker, stem and leaf, histogram)?
How do you know if a set of data is approximately normally distributed? Look it up.
Why do we block?
Why do we experiment?
What makes an experiment special?
What are the characteristics of a well-designed experiment?
Why do people sometimes need double-blind experiments?
What is the placebo effect?
How do you know if two characteristics are independent?
Subscribe to:
Post Comments (Atom)
Posts
1 comment:
Great answers! YOu are going to do well.
The best way to tell if a distribution is normal os to run the normal probability plot. That does the same thing as comparing the 68-95-99.7 points, but it does ALL of them! And it does it quickly.
Good answer on resistant to outliers!
Calm down. See you in a few hours.
Mrs. L
Post a Comment