You are on page 1of 5

Examples of All Kinds of p-values

Note: Decision rules are not explicitly given in the following problems because they are always the same: If
the p-value is less than alpha, reject the null and conclude the alternative; otherwise fail to reject the null.

On October 6, 2012 the Center for Disease Control reported that the rate of cigarette smoking among adults
(persons 18 or older) was 24.1%. In a class survey at the beginning of the semester, 120 persons reported
smoking one or more cigarettes a day out of 517 respondents.

Test One : Is the proportion of IU student smokers in keeping with the rate reported by the CDC?

H 0: = 24.1%
H 1: 24.1%

p (sample proportion) = 120/517 = .232108 or 23.2%


= SQRT((.241*(1-.241))/517) = 0.01881

ZOBS = (p-)/ = (.232-.241)/0.01881 = -0.47847

p-value:
p-value = 2*(1-NORM.S.DIST(ABS(-0.47847)) = 0.632313

Why ABS, 1! and 2*? The ABS in the command lets one ignore the sign on the test statistic and
routinely calculate the area beyond a right tailed test statistic (a positive one.) The 1! is used
because the test statistic is in the right tail (where we forced it to be with the ABS). Finally, the 2* is
used because this is a 2-tailed test (=/); we calculate the area beyond the test statistic and multiply
by 2 to get a value we can compare to . This method will ALWAYS give you the correct p-value for
a two tailed test, regardless of the sign on the test statistic.

Test Results: Since the p-value = 0.632313 or 63.2% is larger than any alpha you would care to
name (for example, 0.01, 0.05, even 0.10) we would fail to reject the null and conclude that the
proportion of adult smokers among the survey respondents is not statistically significantly different
from the population proportion.

Test Two: Suppose the Committee for Showing Off Indiana Students believes that a smaller
proportion of IU students smoke than the national figure contends.

H 0: $ 24.1%
H 1: < 24.1%

p (sample proportion) = 120/517 = 0.232108 or 23.2%


= SQRT(((0.241*(1-0.241))/517) = 0.01881
This is a Left-tailed test.

ZOBS = (p-)/ = (0.232-0.241)/0.01881 = !0.47847

p-value:
p-value = NORM.S.DIST(!0.47847) = 0.316158

Why no ABS, 1! or 2*? No ABS is necessary because we are performing a left-tailed test, and the
standardized test statistic is appropriately negative; there is no need for 1! for the same reason.
Finally, no 2* is necessary because this is a 1-tailed test.

1
Test Results: Since the p-value = 0.316158 or 31.6% is larger than any alpha you would care to
name (for example, 0.01, 0.05, even 0.10) we would fail to reject the null and conclude that the
proportion of adult smokers among the survey respondents is not statistically significantly smaller
than the population proportion.

Test Three: Suppose it was pointed out that 32 of the 517 in our sample survey did not respond
to the question and that the sample proportion was 120/485 or 24.7%. Is the
proportion of IU student smokers in keeping with the rate reported by the CDC?

H 0: = 24.1%
H 1: 24.1%

p (sample proportion) = 120/485 = 0.247423 or 24.7%


= SQRT(((.241*(1-.241))/485) = 0.01942

ZOBS = (p-)/ = (0.247-0.241)/0.01942 = 0.30896

p-value:
p-value = 2*(1-NORM.S.DIST(0.30896) = 0.757352

Why no ABS, but 1! and 2*? ABS can be put into the command just as it was in Test One if you
wish; it will not cause any problem. I left ABS out of the command because it is redundant the test
statistic is positive already. The 1! is used because the test statistic is in the right tail (either because
we used ABS or just because it is positive). Finally, the 2* is used because this is a 2-tailed test
(=/); we calculate the area beyond the test statistic and multiply by 2 to get a value we can compare
to . This method will ALWAYS give you the correct p-value for a two tailed test, IF you calculate
the appropriate standardized test statistic (ZOBS).

Test Results: Since the p-value = 0.757352 or 75.7% is larger than any alpha you would care to
name (for example, 0.01, 0.05, even 0.10) we would fail to reject the null and conclude that the
proportion of adult smokers among the survey respondents is not statistically significantly different
from the population proportion.

Test Four: Suppose the Committee to Bash Smokers believes that a larger proportion of IU students
smoke than the national figure contends.
H 0: # 24.1%
H 1: > 24.1%

p (sample proportion) = 120/485 = 0.247423 or 24.7%


= SQRT(((0.241*(1-0.241))/485) = 0.01942

ZOBS = (p-)/ = (0.247-0.241)/0.01942 = 0.30896


This is a Right tailed test.

p-value:
p-value = 1-NORM.S.DIST(0.30896)) = 0.378676

This is a straight forward application of Excel using the constraints that Excel uses, that it
calculates probability accumulating from the left. A right-tailed test requires us to use 1!; we
dont use 2* because it is a one-tailed test.

Test Results: Since the p-value = 0.378676 or 37.9% is larger than any alpha you would care to
name (for example, 0.01, 0.05, even 0.10) we would fail to reject the null and conclude that the

2
proportion of adult smokers among the survey respondents is not statistically significantly larger than
the population proportion.

Test Five: Just to give you an example of a test where we actually reject a null, I am making a sample
proportion up. I will repeat the last test, using a proportion of 0.278351, or 135 out of 485.
Suppose the Committee to Bash Smokers believes that a larger proportion of IU students
smoke than the national figure contends.
H 0: # 24.1%
H 1: > 24.1%

p (sample proportion) = 120/485 = .278351 or 27.8%


= SQRT(((0.241*(1-0.241))/485) = 0.01942

ZOBS = (p-)/ = (0.278-0.241)/0.01942 = 1.905252


This is a Right tailed test.

p-value:
p-value = 1-NORM.S.DIST(1.905252) = 0.028374

See Test Four on the previous page for a detailed explanation of the command.

Test Results: The p-value = 0.028374 or 2.8%. We would reject the null for any alpha that is
greater than 2.8%, (such as 0.05 or 0.1) and conclude that the proportion of smokers from our survey
is statistically significantly larger than the population proportion. However, we would fail to reject
our null at any alpha smaller than 2.8% (such as .01 or .001) and conclude that the proportion of adult
smokers among the survey respondents is not statistically significantly larger than the population
proportion.

Another Set of All Kinds of p-values: Students t Distribution


Can pleasant aromas help a student learn better? Hirsch and Johnston, of the Smell & Taste
Treatment and Research Foundation, believe that the presence of a floral scent can improve a person's
learning ability in certain situations. In their experiment, 20 people worked through a set of two pencil and
paper mazes six times, three times while wearing a floral-scented mask and three times wearing an unscented
mask. Individuals were randomly assigned to wear the floral mask on either their first three tries or their last
three tries. Participants put on their masks one minute before starting the first trial in each group to minimize
any distracting effect. Subjects recorded whether they found the scent inherently positive, inherently
negative, or if they were indifferent to it. Testers measured the length of time it took subjects to complete
each of the six trials. NOTE: Remember that when working with the Students t, you must use t-scores to get
probabilities(i.e. p-values).
Assume that the population mean time spent for each trial for unscented students is 50.0143 minutes. The
mean of scented students time spent is 49.0571 minutes with standard deviation of 2.3008.

Test One: Test whether the pleasant aromas have any significant effect toward student learning at 3%
level of significance.

Hypotheses :
H0 : = 50.0143
H1 : 50.0143
= 3% = 0.03

This is TWO-TAILED TEST


tOBS = (49.0571-50.0143)/(2.3008/sqrt(60)) = -3.2225

3
p-value =2*(1- T.DIST(ABS(-3.2225),59,1) = 0.00207

Why ABS, 1! and 2*? The ABS in the command lets one ignore the sign on the test statistic and
routinely calculate the area beyond a right tailed test statistic (a positive one.) The 1! is used
because the test statistic is in the right tail (where we forced it to be with the ABS). Finally, the 2* is
used because this is a 2-tailed test (=/); we calculate the area beyond the test statistic and multiply
by 2 to get a value we can compare to . This method will ALWAYS give you the correct p-value for
a two tailed test, regardless of the sign on the test statistic.

Test Results
Since the p-value of 0.00207 is less than 0.03, we reject the null of no difference and conclude that
the pleasant aromas have a significant effect on student learning.
We would begin to fail to reject the null at an alpha of 0.00207.

Conclusion: The pleasant aromas have a significant effect toward student learning at 3% level of
significance.

Test Two: Test whether pleasant aromas help a student learn better (solve maze faster) at 3% level of
significance.

Hypotheses:
H0 : $ 50.0143
H1 : < 50.0143
= 3% = 0.03

This is a LEFT-TAILED TEST

tOBS = (49.0571-50.0143)/(2.3008/sqrt(60)) = -3.2225

p-value = T.DIST(-3.2225,60-1,1) = 0.001035

Test Results
Since the p-value of 0.001035 is less than 0.03, we reject the null of no difference and conclude that
the pleasant aromas have a significant positive effect on student learning.
We would begin to fail to reject the null at an alpha of 0.001035.

Conclusion: Since the average time spent for solving the mazes of those scented students is
significantly shorter than those of unscented students,
pleasant aromas help a student learn better (solve maze
faster) at 3% level of significance.

Test Three: Test whether pleasant aromas worsen a


student learning ability (solve maze slower)
at 3% level of significance.

Hypotheses:
H0 : # 50.0143
H1 : > 50.0143
= 3% = 0.03

This is a RIGHT-TAILED TEST

tOBS = (49.0571-50.0143)/(2.3008/sqrt(60)) = -
3.2225

4
p-value = 1-T.DIST(-3.2225,60-1,1) = 0.998965

Note: This problem demonstrates how important it is to remember that the p-value is the area in the
tail in the direction of the alternative. If the negative sign on the test statistic had been ignored, the
p-value for the test would be identical to that of Test Two above, which would have resulted in
conflicting test results.

Test Results
Since the p-value of 0.998965 is greater than 0.03, we fail to reject the null. There is no evidence
that pleasant aromas cause students to do worse.

Conclusion: Since the average time spent solving the mazes of those scented students is NOT
significantly longer than those of unscented students, pleasant aromas DOES NOT worsen the
students ability to learn (solve maze slower) at 3% level of significance.

You might also like