Statistics and Data Analysis

for Nursing Research

Second Edition

CHAPTER 6

t-Tests: Testing Two Mean Differences

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Two-Sample t-Tests

• Two-sample t-tests are used when the

independent variable is a nominal-level
variable with two levels—that is, when two
groups are being compared
• Examples:
– Experimentals versus controls
– Men versus women
– Children with versus without cystic fibrosis

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Example

• Two groups of patients with Alzheimer’s. One

group received memory-enhancing drug. Other
group received a placebo. Which group can
recall more words from a list 6 months later?
• Is there a difference in words recalled
between durg1 and Placebo Group?
• IV (grouping variable) and DV (outcome
variable).
• Is there a relationship between Drug type and
# of words recalled?
Copyright ©2010 by Pearson Education, Inc.
Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Selecting the appropriate difference
Statistical Test
• Level of Measurement for IV: with difference tests, we
are looking at differences between groups (nominal). We
always have just two groups.

• Level of Measurement for DV (can be interval or ratio).

• Whether groups are dependent or independent!

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Independent Samples

• The subjects selected for one sample have no

influence on the selection of the subjects in the
second sample.

• Example: random assignment to groups

(experimental, control); comparison between
patients with and without cystic fibrosis.

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Dependent Sample

• The subjects selected for one sample

influence or determine the selection of
cases for the other sample.

• EX: pre-test/post-test comparison,

matched group.

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Hypothesis Testing: Example
• Randomly selected 20 elderly individuals who have
Alzheimer’s disease. They are randomly assigned to one
of two groups:

• Experimental: Received an experimental drug for 6

months. Some instructions as control regarding studying
the list of words (group1)

• Control: Received placebo for 6 months. They are given

a memory test at the end of 6 months where they are
instructed to memorize a list of words.
Copyright ©2010 by Pearson Education, Inc.
Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Step 1: State the Null & Alternative
Hypotheses in English Form

• Null hypothesis (H0): How can we write

this up?

• Alternative Hypothesis (H1): what do you

think?

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Step 2: Specify the Test Statistic

# of DV LOM of DV # of PV LOM of PV Categories What test?

LOM for DV Nominal Ordinal Interval/

Ratio
Independent Chi-square Mann Independent
Sample Whitney U sample t-test

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Step 2: Have We Met the Assumptions of the
Test Statistic (Independent Samples t-test)?

Robust to Violation (N ≥ Not Robust to Violation

30)
Random Samples (but Independent Sample
limited generalizability)
Normal Distribution (n’s large DV (Ratio or Interval LOM)
enough and roughly equal)
Homogeneity of Variance

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Assumption: Normal Distribution of DV

• This assumption is robust to violation if n’s

large enough & no major skewness or
kurtosis.
• Ideally, N (n1 +n2) needs to be at least 30.
• If N<30, then you need to make sure n1 is
not > 1.5 times greater than n2.

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Assumption: Homogeneity of Variance

• In order to meet this assumption, it is

recommended that N is greater or equal
30.
• We also look at Levene’s Test for equality
of variances (the alpha level should not be
significant in SPSS output).

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Testing the Assumptions: Our example

• Independent Sample ✅
• Interval/ Ratio LOM for DV ✅
• Random Sample ✅
• Normality? Note: N= 20 ✅ (values of skewness
and kurtosis lied within normal ranges for
experimental and control.
• Homogeneity of Variance ✅
- N ≥ 30? No. We look at Levene’s test in SPSS

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Hypothesis Testing Step 3: Decision
Rule
• Which level of α to select? (you decide)
• Directional or nondirectional (you decide)
• Critical value for the test statistic
• Critical Region
• Decision

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Determining t-Critical Value

Our example:
• df= N-2= 20-2=18
• α= .05.
• Hypothesis: nondirectional.
• By using the t critical values table, tcv= ±2.101.
• Critical region: t≥ 2.101 or t≤ -2.101.

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Raw Data for Our Example

• Dependent Variable (# words recalled)

Group 1 (Placebo) Group 2 (AD New Drug)

12 22 16 15 19 23 32 20 30 24 27
16 14 18 25 30 22 31 25 19
N= 10 N= 10
M= 18 M= 26
St. Dev= 4.216 St. Dev= 4.47

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Decision

• Does our observed t-value fall in the critical

region?
t-value df Sig (two-tailed)

-4.0 18 <.001

• Interpretation of Results: you should report this

results in APA
• t(df)= t-value, p-value.
• t(18)= 4, p<.001.
Copyright ©2010 by Pearson Education, Inc.
Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Our Decision: Reject H0

• If reject H0:
1. Report difference between sample means is
statistically significant
2. Conclude there is a difference between the two
population means
3. Report size of difference (Confidence Intervals
and effect size like d)
4. Discuss possibility of Type one error
5. Strengths/ limitations
Copyright ©2010 by Pearson Education, Inc.
Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
Confidence Intervals and Effect Size

• A CI can be constructed around the difference between

the two group means, to indicate precision of sample
estimate.
• Effect size is a measure of the strength (magnitude) of the
relationship between variables in the population
– When calculated with sample data, an effect size is an
estimate of “how wrong” the null hypothesis is.

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Cohen’s Guidelines

• Cohen provided these qualitative descriptors

for d:
d = .20 = Small effect
d = .50 = Moderate effect
d = .80 = Large effect
Most effect sizes in nursing research are in the
vicinity of .35 (small to moderate)

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Effect Size and Power

• Effect size strongly affects the power of a

statistical test—i.e., the power to correctly
reject the null hypothesis when it is false
• Power is the inverse of β, the risk of a Type
II error
– Power = 1 – β

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Power Analysis

• Power analysis is most often used during the

planning phase of a study to estimate how many
participants are needed to minimize the risk of a
Type II error
• Just as .05 is the standard acceptable risk for a
Type I error, .20 is the standard for a Type II error
– So, minimum acceptable power = .80

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 Power Analysis Components
• There are four components to a power
analysis—three must be known or
estimated to solve for the fourth:
1. Significance criterion (.05)
2. Power (.80)
3. Population effect size
4. Sample size

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 t-Tests in SPSS:
Independent Groups
• Select Analyze →
Compare Means →
Independent Samples
T Test

• Move dependent
variable from list into
slot for Test Variables
Copyright ©2010 by Pearson Education, Inc.
Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 t-Tests in SPSS:
Independent Groups (cont’d)

• Move independent
variable into slot for
Grouping Variable
• Click Define Groups
pushbutton

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 t-Tests in SPSS:
Independent Groups (cont’d)

• In the next dialog

box, enter the
values associated
with the two groups
being compared

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.
 t-Tests in SPSS:
Dependent Groups
• Select Analyze → Compare Means → Paired
Samples T Test
• Select one or more pairs of variables, moving
them from list into Paired Variables slot

Copyright ©2010 by Pearson Education, Inc.

Statistics and Data Analysis for Nursing Research, Second Edition
Upper Saddle River, New Jersey 07458
Denise F. Polit
All rights reserved.