Professional Documents
Culture Documents
the sum = 0) Constant std dev of errors (check the residual plot for uniform, random scatter) Normal errors (check the normal probability plot for linearity) Reading Computer Output:
Regression output t (df=10)
variables
coefficients
std. error
p-value
Intercept Age
64.9283 0.6350
0.5084 0.0214
127.709 29.665
2.13E-17 4.43E-11
S=0.256
R-sq = 98.9%
R-sq(adj)
= 97.2%
Standard error of the line Standard error of the slope Remember that the p-value is a two sided p-value, so divide by 2 of you want a one sided p-value. But dont forget that t = b/SEb (in case you are asked to find it!) And to find a confidence interval: b t*SEb Interpretations: : The change in the mean y for each 1 unit increase in x. s: Tells you how much the observed y-values differ from the predicted y-value, for a given x value. SEb: In repeated sampling, how much the slopes of LSRLs would differ from the true regression slope . When interpreting your confidence interval, include the interpretation of the slope. For example, I am 95% confident that the mean height increases by between 0.2 and 0.7 cm per month. Helpful Tips: Know where the formula for SEb is on your formula sheet and recall that the numerator is the same as s. A time saving tip, once you find the residuals, just do 1-var stat to get the sum of the squares of the x2 residuals
Instead of doing linear regression in your calculator to get the LSRL, if you know you will be doing a Linear Regression T-Test, go ahead and do that since it will give you t, p, etc AND the LSRL.