Understanding Point Estimates, Confidence Intervals, and Hypothesis Testing

Test 4

Point estimate – single number computed from a sample, which serves as a guess for the parameter.

Estimator – statistic of interest and is therefore a random variable.

Estimate – simply a specific value of an estimator.

Confidence interval – Interval of values constructed so that, with a specified degree of confidence, the value of the population parameter lies within it.

Confidence coefficient – Probability that the CI includes the population parameter in repeated samples.

Confidence level – The confidence coefficient expressed as a percentage.

Confidence interval example when SD is known

Proportion Test Example (If p<= a: Reject null, If p>a: dont)

We need a 95% CI for μ. The population is normal and SD is known. n=2500, a=0.01, x=935

x̄ = 155.7; SD = 18, n = 15

p-x/n=935/2500=0.3740

1 – a = 0.95 -> a = 0.05 -> a/2 = 0.025

Ho: p=0.36, Ha: p>0.36, TS: Z=(P-po)/(SQRT(po(1-po)/n))

P(Z≥z0.025)=1-0.025=0.975

RR: Z>=za=z0.01=2.3263

z0.025=1.96

Ho: p=0.36, Ha: p>0.36, TS: Z=(P-po)/(SQRT(po(1-po)/n)

x ± za/2*SD/sqrt(n)

Another example (n=400, x=204, po=0.58, a=0.05)

=x ± z0.025*SD/sqrt(n)

P=x/n=204/400-0.51

=(146.59,164.81)

Ho: p=0.58

Bound of Error of Estimation Example

Ha: p != 0.58

SD=15, B=5

RR: |Z|>=za/2=z0.025=1.96

1-a=0.95->a/2=0.025->z0.025=1.96

((SD* za/2)/B)^2 = 34.57 = n>=35 (round up)

normcdf(-1e99,-2.84,1,0) = 0.0023

Confidence interval example when SD is unknown

p=2(0.0023)=0.0046(<=0.05=a) we reject

x̄ = 980.2; SD = 27.6, n = 11

1-a=0.95->a=0.05->a/2=0.025

Find t critical value

ta/2,n-1=t0.025,10=2.2281

n=15.95%

x ± ta/2,n-1*s/sqrt(n)

CI for p and inference (p=x/n)

invT(a/2,n-1)

We need a 95% CI for p. If it asks if P is normal test if its greater

n=1200 than 5. (n)(P)>=5 and (n)(1-P)>=5

1-a=0.95->a=0.05->a/2=0.025

za/2=z0.025=1.96

p± za/2*SQRT(p(1-p)/n)

Error of Estimation

B=za/2*SQRT(p(1-p)/n)

n=p(1-p)(za/2/B)^2

Hypothesis Test Errors

Reject Ho Do not reject Ho Ho: u = uo

Truth Ho Type 1 error Correct Ha: u > uo, u < uo, or u != uo

Ha Correct Type II error TS: Z=(X-uo)/(SD/sqrt(n))

RR: Z>=za, Z<=za, or |Z|>=za/2

P value example (u=670;n=20;a=25)

Ho: u=670,Ha: u<670,TS: Z=(x-uo)/(a/sqrt(n)) = -1.771 Random

p=P(Z<=-1.771)=P(Z<=-1.77)=0.0384 Sx = population variance. Make sure to ^2 it

because p=0.0384<=0.05 we reject Ho. Find critical value. Ex. z0.10 InvNorm(0.10,0,1) <– make positive

two sided example

p/2=P(Z>=1.34) = 0.0901 = p=2(0.0901)=p=0.1802