Final Exam Tips

Compute p-value for a lower tail test on the population mean: =t.dist(t-stat, d.f., 1)

Fixing everything else, margin of error increases in confidence level

Confidence Interval:
point estimate +/- (critical value) * (standard error)

Margin of Error:
critical value * standard error

Three layers of analysis in regression:
1) overall model
2) residual analysis (regression assumptions)
3) individual coefficients

Insignificant independent variables need to be dropped in the regression analysis

If sample size n < 30, you can't construct a 95% confidence interval for population proportion

In regression, approximated 95% prediction interval:
point estimate +/- 2 * Se

Critical t-value can be obtained via:
=t.inv.2t(probability of both tails, d.f.)

Critical z-value corresponds to:
90% confidence level: 1.645
95% confidence level: 1.96
99% confidence level: 2.576

Two things to notice when checking residual plots:
1) Spread should be constant (no heteroskedasticity)
2) There should be no increasing, decreasing, or cyclic trends (no autocorrelation)

1) increase in sample size = decrease in margin of error
2) increase in stdev = increase in margin of error
3) increase in condidence interval = increase in margin of error

Type 1 Error: Occurs when null hyp. is true and you reject null hyp.
Solve: Reduce the sig level (arva)

Type 2 Error: Occurs when null hyp. is false and you don't reject null hyp.
Solve: Increase sample size

In regression, rejection is always a good thing

Every sample contains sampling error

1) want higher adjusted R2
2) want low stderror of estimates
3) want low P-value for f test