Documente Academic
Documente Profesional
Documente Cultură
• Matched pairs: To compare the responses to the • Large-sample z confidence interval for p1 − p2 :
two treatments, apply the one-sample t proce-
dures to the observed differences. (p̂1 − p̂2 ) ± z ∗ SE z ∗ from N (0, 1)
• Two-sample t confidence interval for µ1 − µ2 (in- where the standard error of p̂1 − p̂2 is
dependent SRSs from Normal populations): s
s p̂1 (1 − p̂1 ) p̂2 (1 − p̂2 )
SE = +
s21 s22 n1 n2
(x1 − x2 ) ± t∗ +
n1 n2
Plus four to greatly improve accuracy: use the
with conservative t∗ from t with df the smaller same formulas after adding one success and one
of n1 − 1 and n2 − 1 (or use software). failure to each sample.
• Two-sample z test statistic for H0 : p1 = p2 • t confidence interval for regression slope β:
(large independent SRSs):
b ± t∗ SEb t∗ from t(n − 2)
p̂1 − p̂2
z=s
1 1 • t test statistic for no linear relationship,
p̂(1 − p̂) +
n1 n2 H0 : β = 0:
• The regression model: We have n observations • Describe the data using the I sample means and
on x and y. The response y for any fixed x has standard deviations and side-by-side graphs of
a Normal distribution with mean given by the the samples.
true regression line µy = α + βx and standard
• The ANOVA F test statistic (use software) is
deviation σ. Parameters are α, β, σ.
F = MSG/MSE, where
• Estimate α by the intercept a and β by the slope
n1 (x1 − x)2 + · · · + nI (xI − x)2
b of the least-squares line. Estimate σ by the MSG =
regression standard error: I −1
(n1 − 1)s1 + · · · + (nI − 1)s2I
2
s MSE =
1 X N −I
s= residual2
n−2