Statistics and Data Analysis for

Nursing Research
Second Edition

Chapter 6
t-Tests: Testing Two Mean
Differences

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Two-Sample t-Tests (1 of 2)
• 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 Pearson Education, Inc. All Rights Reserved

Two-Sample t-Tests (2 of 2)
• Two-sample t-tests are used when the dependent variable
is an interval- or ratio-level variable—that is, a variable for
which it is appropriate to compute a mean
• Examples:
– Weight (pounds or kilos)
– Scores on a stress scale
– Fatigue (measured on a 100-point visual analog scale)

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Null Hypotheses for t-Tests
• The null hypothesis is that the two group means are the
same—i.e., that the independent variable and dependent
variable are not related
• Formally stated:
– H 0 :μ1 = μ 2 where
– μ1 = the mean for Group 1 and
– μ 2 = the mean for Group 2

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Alternative Hypotheses for t-Tests
• The alternative (research) hypothesis is that the two group
means are different—that is, that the independent variable
and dependent variable are related
• Formally stated (nondirectional):
– H 0 :μ1  μ 2 where
– μ1 = the mean for Group 1 and
– μ 2 = the mean for Group 2

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Testing the Null Hypothesis
• In a two-sample t-test, testing the null hypothesis involves
computing a test statistic and comparing it to values of
what is “improbable,” if the null hypothesis were true
– What is “improbable” lies in the tails of a theoretical
sampling distribution
• Here, the relevant distribution is the sampling distribution
of the difference between two means

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Sampling Distribution
• The mean of this sampling distribution of the difference
between two means always equals the population mean
difference
• The standard error of this distribution is the standard error
of the difference (SE D ) and is a key component in
computing the t statistic

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Assumptions for t-Tests (1 of 2)
• To use a t-test, it is assumed that:
– Cases have been independently and randomly sampled
from the two populations
– The outcome variable is normally distributed in the two
populations
– The two population variances are equal: This is the
assumption of the homogeneity of variance

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Assumptions for t-Tests (2 of 2)
• Robustness of the t-test—violations of the last two
assumptions do not affect statistical decision making if:
– Sample sizes are large (40 + per group)
– Sample sizes in the two groups are similar (less than
1.5 times the number in one group as in the other
group)

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Tests for Assumptions
• With small samples, or severely unequal group sizes,
violation of assumptions should be tested
• Test for normality: the Kolmogorov-Smirnov test
• Test for equal variances: Levene’s F test

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Types of t-Tests
• There are three different formulas for computing a t
statistic, all of which share the goal of testing mean
differences between two groups:
– Independent groups t-test
▪ Pooled variance formula
▪ Separate variance formula
– Dependent groups t-test

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Independent Groups t-Tests
• Independent groups t-test: Used to test the difference
between means for two unrelated groups
• Examples:
– Men versus women
– Teenagers versus adults
– Lung cancer patients versus pancreatic cancer patients
• The people in the two groups are not the same people,
and they are not related to each other in any systematic
way

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Pooled Variance Formula (1 of 2)
• The basic formula for the independent groups t-test uses a
pooled variance estimate for the standard error of the
difference
– The numerator in the equation to compute t is the
difference in sample means ( M 1 − M 2 )
– The denominator is the estimated SED
– Degrees of freedom = n1 + n2 − 2
• If calculated t (absolute value)  tabled t (for appropriate
df and ) the result is statistically significant

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Pooled Variance Formula (2 of 2)
• The pooled variance formula can be used if the population
variances are equal–tested via Levene’s test
– If the F from this test is significant, the pooled variance
formula should not be used
• Levene’s test is automatically run within SPSS when
doing an independent groups t-test

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Levene’s Test
• Tests the null hypothesis that the variances in the two
populations are equal
• SPSS printout below: The null hypothesis is retained,
p = .719
– Thus, it’s OK to use the pooled variance estimate

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Separate Variance Formula
• If Levene’s test indicates statistically significant
differences in variances, the separate variance formula
should be used to estimate SED in computing the t statistic

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

t- Test Printout (1 of 2)
• SPSS printout for independent groups t-test
• Same result for both formulas
– Means are not significantly different when  = .05,
two-tailed test: p = .081

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Dependent Groups t-Tests (1 of 2)
• Dependent groups t-test: Used to test the difference
between means for two related groups, or for the same
people measured twice
– Also called a paired t-test or correlated groups t-test
• Examples:
▪ Preintervention versus postintervention
▪ Husbands and wives
▪ Twins
• The people in the two groups are systematically
connected, which lowers variability (the SED )

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Dependent Groups t-Tests (2 of 2)
• Dependent groups t-test formula for computing t :
– Numerator: The mean difference between pairs of
values ( M D )
– Denominator: Estimated standard error of the
difference
– df = number of pairs − 1

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

t- Test Printout (2 of 2)
• SPSS printout (selected segments) for dependent groups t-test
• Mean differences in direct and indirect measurement of SBP for
the same people are not statistically significant
( p = .090) → Null hypothesis is retained

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Other Statistical Issues
• Existence of a relationship: t-tests
• Nature of relationship: Direct comparison of the two
means
• Precision of estimates: Confidence intervals around the
mean difference
• Magnitude of estimates: Effect size

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Confidence Intervals
• A CI can be constructed around the difference between
the two group means, to indicate precision of sample
estimate
• Formula for CIs:
– Compute the mean group difference ( M 1 − M 2 )
– Then add/subtract SED  the tabled value of t for correct
CI (e.g., 95%) and df

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Effect Size
• 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
• Effect size is a critical construct in meta-analyses—
analyses that integrate the results of multiple studies on a
given question statistically

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Cohen’s d
• In a two-group mean-difference situation (like when a t-test is
called for), the effect size is Cohen’s d
– Sometimes called the standardized mean difference (SMD)
• The d statistic captures the amount of difference in the two
means, in standard deviation units
– d = ( M 1 − M 2 )  pooled SD

Copyright © 2010 Pearson Education, Inc. 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 Pearson Education, Inc. 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  the risk of a Type  error
– Power =  − 

One fails to reject a null hypothesis that is actually false. A type II error (false
negative)

Copyright © 2010 Pearson Education, Inc. 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  error
• Just as .05 is the standard acceptable risk for a Type 
error, .20 is the standard for a Type  error
– So, minimum acceptable power = 

Copyright © 2010 Pearson Education, Inc. 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 Pearson Education, Inc. All Rights Reserved

Sample Size Estimation (1 of 2)
• To estimate the sample size needed to reduce the risk of
a Type  error to .05 and the risk of a Type  error to .20,
you need an estimate of effect size (d in a two-group
mean difference situation)

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Sample Size Estimation (2 of 2)
• Effect size estimates:
– Prior similar studies
– A pilot study
– Cohen’s guidelines (assume no more than small-to-
moderate unless there is evidence to the contrary)
– Minimal effect that would be cost effective and/or
clinically relevant, and that seems feasible

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Power Analysis Examples
• Assume  =  and  = 
• Sample size needs (per group) for different ds:
– d = .20 (small effect): 394/group
– d = .35 (small-to-moderate effect): 129/group
– d = .50 (moderate effect): 64/group
– d = .80 (large effect): 25/group
• When group ns are not equal, samples must be larger, all
else equal

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Post Hoc Power Analysis
• It is sometimes useful to estimate what power actually was
in a study, especially if results are not significant
– Somewhat controversial, but can be useful in
interpretation and in thinking about “next steps” for
moving an area of research forward
• Here, effect size and sample size are known, and so the
analysis solves for power

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Post Hoc Power Analysis: Some
Examples
• Assume  = 
•  = Risk of committing a Type  error:

n (per group) d Power 

Beta

25 .20 (small) .11 .89

25 .50 (moderate) .42 .58
25 .80 (large) .81 .19
50 .20 (small) .17 .83
50 .50 (moderate) .71 .29
50 .80 (large) .98 .02

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Uses of t-Tests (1 of 4)
• Primary use of t-tests—Substantive:
– To answer research questions and test substantive
research hypotheses involving comparisons between
two groups
• Other uses—Methodologic

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Uses of t-Tests (2 of 4)
• Evaluations of a study’s internal validity through bias
assessment
– Internal validity concerns the strength of inferences
about the degree to which the independent variable
really affected the dependent variable

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Uses of t-Tests (3 of 4)
• Bias assessment examples:
– Selection bias: Are groups being compared really
comparable (e.g., experimentals and controls)?
– Attrition bias: Are people who dropped out similar to
those who stayed in a study?

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Uses of t-Tests (4 of 4)
• To contribute evidence about the construct validity of a
scale, through the use of the known-groups technique
– Involves comparing two groups who are hypothesized,
based on theory or prior research, to differ on the new
instrument

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

t-Tests in SPSS: Independent Groups (1 of 3)

• Select Analyze
Compare Means
Independent Samples
T Test
• Move dependentAn arrow pointing to dependent variable.

variable from list into

slot for Test Variables
An arrow pointing to Test Variables.

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

t-Tests in SPSS: Independent Groups (2 of 3)

• Move independent
variable into slot for
Grouping Variable An arrow pointing to Grouping Variable.

• Click Define Groups

pushbutton An arrow pointing to Define Groups.

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

t-Tests in SPSS: Independent Groups (3 of 3)

• In the next dialog

box, enter the
values associatedAn arrow pointing to use specified values: Group 2.
1.

with the two groups

being compared

Copyright © 2010 Pearson Education, Inc. 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
An arrow pointing to paired variables slot.

Copyright © 2010 Pearson Education, Inc. All Rights Reserved

Copyright

This work is protected by United States copyright laws and is

provided solely for the use of instructors in teaching their
courses and assessing student learning. Dissemination or sale
of any part of this work (including on the World Wide Web) will
destroy the integrity of the work and is not permitted. The work
and materials from it should never be made available to students
except by instructors using the accompanying text in their
classes. All recipients of this work are expected to abide by
these restrictions and to honor the intended pedagogical
purposes and the needs of other instructors who rely on these
materials.

Copyright © 2010 Pearson Education, Inc. All Rights Reserved