You are on page 1of 2

Type I Error: Tests of One Mean

Test statistic:
Objective To illustrate hypothesis Scenario The Dean of Students says that Faber H0: m £ 2.50 Press F9 to take
tests for a mean and let you study College's 1,000 transfer students maintain average samples from this x  0
Press F9 to take samples H1: m > 2.50 tcalc 
Type I error empirically. GPAs at or below 2.50. Do these recent samples population. s
from this population. support the Dean's belief?
n

N = 1000 n=5 n = 10 n = 20 Sample Statistics


Population Small Sample Medium Sample Large Sample Statistic n=5 n=10 n=20
Obs GPA Row GPA Row GPA Row GPA Mean 2.402 2.306 2.590
1 2.47 940 1.79 614 2.11 131 2.92 St Deviation 0.395 0.387 0.396
2 2.39 248 2.66 238 2.46 396 2.48 Count 5 10 20
3 2.56 184 2.43 856 2.98 171 2.55
4 2.69 230 2.82 725 2.79 881 2.61 Right-Tailed Hypothesis Test
5 2.63 97 2.31 685 2.11 322 1.97 Statistic n=5 n=10 n=20
6 2.31 374 1.83 963 1.97 d.f. 4 9 19
7 3.03 920 2.26 830 2.96 t-statistic -0.554 -1.586 1.010
8 3.06 24 1.77 628 3.72 p-value 0.696 0.926 0.163
9 2.16 913 2.50 509 2.36 Critical t.05 -1.533 -1.383 -1.328
10 2.95 125 2.25 271 2.56 Reject H0? Reject No Reject
11 2.68 740 2.72
12 2.88 Excel Tip This example uses simple random sampling 384 2.15
13 2.35 to ensure that every population item has the same 122 2.85
14 2.79
chance of being selected. Study the formulas in the 165 2.12 -1.328
cells to see how the random sample is drawn using n=20
15 2.25 969 2.57 1.010
Excel functions =RAND() and =INDEX.
16 2.91 295 2.87
17 2.45 141 2.56 -1.383 Critical t
n=10
18 2.81 277 2.45 -1.586 Sample t
19 2.12 Note Population info is provided 690 2.70
20 3.12 here (but in real life, you would 269 2.70 -1.533
21 2.50 only see the sample). n=5
-0.554
22 2.56
23 2.23 -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000
24 1.77
25 2.35 True Population Parameters
Population Histogram (N=1000)
26 2.34 Mean St. Dev. N Questions (1) For n = 5, watch the decision as you press F9 twenty times.
27 2.36 2.491 0.355 1000 How many times did you reject H0 at a = 0.05? (2) Since H0 is true (see
500
28 2.98 true population mean) at a = 0.05 you would expect to reject H0 (and
Frequency

400
29 2.60 300 commit Type I error) about 1 time in 20. Did that happen? (3) Repeat
30 2.19 200 using n = 10 and n = 20. (4) Why doesn't sample size affect the
31 2.16 100 probability of Type I error? (5) Why is the p-value more versatile than
32 2.33 0
using a = 0.05?
33 2.55 00 50 00 50 00 50
1. 1. 2. 2. 3. 3.
34 2.19
35 2.15
36 2.49
37 2.55
38 2.31 Note: Print area is truncated due to large population.
39 3.06
40 2.63
41 1.94
42 1.95
43 2.85
Type I Error: Tests of One Proportion
Note Population info is provided
Objective To illustrate Scenario The Chief of Medicine claims that no here (but in real life, you would
Press F9 to take samples hypothesis tests for a more than than 30% of the emergency arrivals are H0: p £ 0.30 only see the sample).
from this population. proportion. on Medicare. Do these samples taken from 1,000 H1: p > 0.30
recent arrivals support that view?

Population of Sample of Sample of Population


1000 Patients 25 Patients 50 Patients Sample Results Admits 300
Patient Medicare? Patient Medicare? Patient Medicare? n=20 n=50 Patients 1000
1 No 450 Yes 711 Yes Admits 6 21 Proportion 0.300
2 No 961 Yes 208 Yes Patients 20 50
3 No 831 No 273 No Proportion 0.30 0.42 Questions (1) For n = 20, observe the decision as you press F9 twenty
4 No 838 No 691 No times. How many times did you reject H0 at a = 0.05? (2) Since H0 is
5 Yes 326 No 450 No Right-Tailed Hypothesis Test true (see true population proportion) at a = 0.05 you would expect to
6 No 357 Yes 921 Yes n=25 n=50 reject H0 (and commit Type I error) about 1 time in 20. Did that
7 No 184 No 962 No z-statistic 0.000 1.852 happen? (3) Repeat using n = 50. (4) Why doesn't sample size affect
8 No 959 No 224 No p-value 0.500 0.032 the probability of Type I error? (5) Is n = 20 large enough to give a
9 No 914 No 800 No Critical z.05 1.645 1.645 reliable test? (6) Why is the p-value more versatile than using a =
10 No 43 No 30 No Reject H0? No Reject 0.05?
11 Yes 203 Yes 716 Yes
12 No 534 No 311 No Test statistic:
13 No 647 No 27 No p 0
14 Yes 398 No 918 No zcalc 
15 No 378 No 540 No
1.645  0 (1   0 )
n=50
16 Yes 46 Yes 2 Yes n
17 No 922 No 649 No Critical z
18 No 488 No 400 No Sample z
19 Yes 447 No 843 No 1.645
20 No 304 Yes 200 Yes n=25
21 No 227 Yes 828 Yes
22 No 201 Yes 109 Yes
23 No 339 Yes 11 Yes -3.000 -2.000 -1.000 0.000 1.000 2.000 3.000
24 No 647 No 384 No
25 No 845 No 156 No
26 Yes 751 Yes Excel Tip This example uses simple random sampling
27 Yes 278 Yes to ensure that every population item has the same
28 No 635 No chance of being selected. Study the formulas in the
29 Yes 310 Yes cells to see how the random sample is drawn using Note: Print area is truncated due to
30 No 913 No Excel functions =RAND() and =INDEX. large population.
31 Yes 863 Yes

You might also like