PASS Sample Size Software NCSS.

com

Chapter 601

Hotellings One-
2
Sample T
Introduction
This module calculates power for Hotellings one-group T-squared (T2) test statistic. Hotellings One-Sample T2
is an extension of the univariate one-sample T-test to the case where the number of response variables is greater
than one.

Assumptions
The following assumptions are made when using Hotellings T2 to analyze one group of data.
1. The response variables are continuous.
2. The residuals follow the multivariate normal probability distribution with mean zero and constant
variance-covariance matrix.
3. The subjects are independent.

Technical Details
The formulas used to perform a Hotellings T2 power analysis provide exact answers if the above assumptions are
met. These formulas can be found in many places. We use the results in Rencher (1998). We refer you to that
reference for more details.

One-Group Technical Details

In one-group case, a set of N observations is available on p response variables. We assume that all N observations
have the same multivariate normal distribution with mean vector and variance covariance matrix and that
Hotellings T2 is used for testing the null hypothesis that = 0 versus the alternative that = A where at least
one component of A is different from the corresponding component of 0 . Usually, the vector 0 is a vector
of zeros.
The value of T2 is computed using the formula

Tp2,N 1 = N ( y 0 ) S 1 ( y 0 )
where y is the vector of sample means and S is the sample variance-covariance matrix.
To calculate power we need the non-centrality parameter for this distribution. This non-centrality parameter is
defined as follows

601-1
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

= N ( A 0 ) 1 ( A 0 )
= N2
where

= ( A 0 ) 1 ( A 0 )
We define as effect size because it provides a expression for the magnitude of the standardized difference
between the null and alternative means.
Using this non-centrality parameter, the power of the Hotellings T2 may be calculated for any value of the means
and standard deviations. Since there is a simple relationship between the non-central T2 and the non-central F,
calculations are actually based on the non-central F using the formula

= Pr( F < F,df 1,df 2, )

where
df 1 = p
df 2 = N p

Procedure Options
This section describes the options that are specific to this procedure. These are located on the Design and
Covariance tabs. For more information about the options of other tabs, go to the Procedure Window chapter.

Design Tab
The Design tab contains many of the options that you will be primarily concerned with.

Solve For
Solve For
This option specifies the parameter to be solved for.
When you choose to solve for Sample Size, the program searches for the lowest sample size that meets the alpha
and power criterion you have specified.

Power and Alpha

Power
This option specifies one or more values for power. Power is the probability of rejecting a false null hypothesis,
and is equal to one minus Beta. Beta is the probability of a type-II error, which occurs when a false null
hypothesis is not rejected. In this procedure, a type-II error occurs when you fail to reject the null hypothesis of
equal means when in fact the means are different.
Values must be between zero and one. Historically, the value of 0.80 (Beta = 0.20) was used for power. Now,
0.90 (Beta = 0.10) is also commonly used.
A single value may be entered here or a range of values such as 0.8 to 0.95 by 0.05 may be entered.

601-2
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

Alpha
This option specifies one or more values for the probability of a type-I error. A type-I error occurs when a true
null hypothesis is rejected. In this procedure, a type-I error occurs when you reject the null hypothesis of equal
means when in fact the means are equal.
Values must be between zero and one. Historically, the value of 0.05 has been used for alpha. This means that
about one test in twenty will falsely reject the null hypothesis. You should pick a value for alpha that represents
the risk of a type-I error you are willing to take in your experimental situation.
You may enter a range of values such as 0.01 0.05 0.10 or 0.01 to 0.10 by 0.01.

Sample Size
Sample Size
This is the number of individuals measures (multiple times).

Effect Size Response Variables

Number of Response Variables
Enter the number of response (dependent or Y) variables. For a true multivariate test, this value will be greater
than one.
The number of mean differences entered in the Mean Differences box or in the Means column must equal this
value. If you read-in the covariance matrix from the spreadsheet, the number of columns specified must equal this
value.

Effect Size Mean Differences

Mean Differences (= # of Response Vars)
Enter a list of values representing the mean differences under the alternative hypothesis. Under the null
hypothesis, these values are all zero. The values entered here represent the differences that you want the
experiment (study) to be able to detect.
Note that the number of values must match the number of Response Variables.
If you like, you can enter these values in a column on the spreadsheet. This column is specified using the Means
Column option. When that option is specified, any values entered here are ignored.
Means Differences Column
Use this option to specify the spreadsheet column containing the hypothesized mean differences. The response
variables are represented down the rows. The number of rows with data must equal the number of response
variables. When this option is used, the 'Mean Differences' box is ignored.
You can obtain the spreadsheet by selecting Window, then Data, from the menus.

Effect Size Mean Multiplier

K (Means Multipliers)
These values are multiplied times the mean differences to give you various effect sizes. A separate power
calculation is generated for each value of K. If you want to ignore this setting, enter 1.

601-3
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

Covariance Tab
This tab specifies the covariance matrix.

Covariance Matrix Specification

Specify Which Covariance Matrix Input Method to Use
This option specifies which method will be used to define the covariance matrix.

Standard Deviation and Correlation

This option generates a covariance matrix based on the settings for the standard deviation (SD) and the pattern
of correlations as specified in the Correlation Pattern and R options.

Covariance Matrix Variables

When this option is selected, the covariance matrix is read in from the columns of the spreadsheet. This is the
most flexible method, but specifying a covariance matrix is tedious. You will usually only use this method
when a specific covariance is given to you.
Note that the spreadsheet is shown by selecting the menus: Window and then Data.

Covariance Matrix Specification- Input Method =

Standard Deviation and Correlation
The parameters in this section provide a flexible way to specify , the covariance matrix. Because the covariance
matrix is symmetric, it can be represented as
11 12 1 p
22 2p
=
12

1 p 2 p pp
12 1 2 12 1 p 1 p

22 2 p 2 p
= 1 2 12


1 p 1 p 2 p 2 p p
2

1 0 0 1 12 1 p 1 0 0
0 0 12 1 2 p 0 2 0
=
2

0 0 p 1 p 2 p 1 0 0 p
where p is the number of response variables.
Thus, the covariance matrix can be represented with complete generality by specifying the standard deviations
1 , 2 ,, p and the correlation matrix

601-4
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

1 12 1 p
1 2 p
R=
12
.


1 p 2 p 1

SD (Common Standard Deviation)

This value is used to generate the covariance matrix. This option specifies a single standard deviation to be used
for all response variables. The square of this value becomes the diagonal elements of the covariance matrix. Since
this is a standard deviation, it must be greater than zero.
This option is only used when the first Covariance Matrix Input Method is selected.
R (Correlation)
Specify a correlation to be used in calculating the off-diagonal elements of the covariance matrix. Since this is a
correlation, it must be between -1 and 1. This option is only used when the first Covariance Matrix Input Method
is selected.
Specify Correlation Pattern
This option specifies the pattern of the correlations in the variance-covariance matrix. Two options are available:

Constant
The value of R is used as the constant correlation. For example, if R = 0.6 and p = 6, the correlation matrix
would appear as
1 0.600 0.600 0.600 0.600 0.600
0.600 1 0.600 0.600 0.600 0.600

0.600 0.600 1 0.600 0.600 0.600
R=
0.600 0.600 0.600 1 0.600 0.600
0.600 0.600 0.600 0.600 1 0.600

0.600 0.600 0.600 0.600 0.600 1

1st-Order Autocorrelation
The value of R is used as the base autocorrelation in a first-order, serial correlation pattern. For example, R =
0.6 and p = 6, the correlation matrix would appear as
1 0.600 0.360 0.216 0130
. 0.078
0.600 1 0.600 0.360 0.216 .
0130

0.360 0.600 1 0.600 0.360 0.216
R=
0.216 0.360 0.600 1 0.600 0.360
0130
. 0.216 0.360 0.600 1 0.600

0.078 0130
. 0.216 0.360 0.600 1
This pattern is often chosen as the most realistic when little is known about the correlation pattern and the
responses variables are measured across time.

601-5
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

Covariance Matrix Specification- Input Method =

Covariance Matrix Variables
This option instructs the program to read the covariance matrix from the spreadsheet.
Spreadsheet Columns Containing the Covariance Matrix
This option designates the columns on the current spreadsheet holding the covariance matrix. It is used when the
Specify Which Covariance Matrix Input Method to Use option is set to Covariance Matrix Variables. The
number of columns and number of rows must match the number of response variable at which the subjects are
measured.

Example 1 Power and Validation

Rencher (1998) page 106 presents an example of power calculations for the one-group case in which the mean
differences are both 1.88 and the covariance matrix is
56.78 1198
.
=
1198
. 29.28
When N is 25 and the significance level is 0.05, Rencher calculated the power to be 0.3397.
To allow for a nice chart, we will calculate the power for several samples sizes and for K equal 1.0 and 1.5.

Setup
This section presents the values of each of the parameters needed to run this example. First, from the PASS Home
window, load the Hotellings One-Sample T2 procedure window by expanding Means, then clicking on
Multivariate Means, and then clicking on Hotellings One-Sample T2. You may then make the appropriate
entries as listed below, or open Example 1 by going to the File menu and choosing Open Example Template.
You can see that the values have been loaded into the spreadsheet by clicking on the spreadsheet button.
Option Value
Design Tab
Solve For ................................................ Power
Alpha ....................................................... 0.05
Sample Size ............................................ 5 15 25 35 50 75 100 150
Number of Response Variables .............. 2
Mean Differences.................................... 1.88 1.88
Mean Differences Column ...................... blank
K (Means Multiplier) ................................ 1.0 1.5
Covariance Tab
Specify Covariance Method .................... Covariance Matrix Columns
Spreadsheet Columns ............................ VC1-VC2
Press the Spreadsheet button to enter the following values into the spreadsheet for columns VC1 and VC2
Row 1 ...................................................... 56.78 11.98
Row 2 ...................................................... 11.98 29.28
Reports Tab
Show Numeric Reports ........................... Checked
Show Means Matrix ................................ Checked
Show Covariance Matrix ......................... Checked

601-6
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

Annotated Output
Click the Calculate button to perform the calculations and generate the following output.

Numeric Report
Multiply Number
Means Effect of Y's
Power N By Alpha Beta Size (DF1) DF2
0.0737 5 1.0000 0.0500 0.9263 0.38 2 3
0.1996 15 1.0000 0.0500 0.8004 0.38 2 13
0.3379 25 1.0000 0.0500 0.6621 0.38 2 23
0.4707 35 1.0000 0.0500 0.5293 0.38 2 33
0.6409 50 1.0000 0.0500 0.3591 0.38 2 48
0.8311 75 1.0000 0.0500 0.1689 0.38 2 73
0.9282 100 1.0000 0.0500 0.0718 0.38 2 98
0.9895 150 1.0000 0.0500 0.0105 0.38 2 148
0.1040 5 1.5000 0.0500 0.8960 0.38 2 3
0.4033 15 1.5000 0.0500 0.5967 0.38 2 13
0.6635 25 1.5000 0.0500 0.3365 0.38 2 23
0.8302 35 1.5000 0.0500 0.1698 0.38 2 33
0.9475 50 1.5000 0.0500 0.0525 0.38 2 48
0.9943 75 1.5000 0.0500 0.0057 0.38 2 73
0.9995 100 1.5000 0.0500 0.0005 0.38 2 98
1.0000 150 1.5000 0.0500 0.0000 0.38 2 148

Report Definitions
Power is the probability of rejecting a false null hypothesis. Note that Power = 1 - Beta.
N is the sample size, the number of subjects in the experiment or study.
K is a constant by which all means are multiplied.
Alpha is the probability of rejecting a true null hypothesis.
Beta is the probability of accepting a false null hypothesis. Note that Beta = 1 - Power.
Effect Size is a standardized version of T2 under the alternative hypothesis.
DF1 is the first degrees of freedom of T2. It is the number of response variables.
DF2 is the second degrees of freedom of T2.

Summary Statements
A sample size of 5 achieves 7% power to detect an effect size of 0.38 which represents the
differences between the null and alternative means of the 2 response variables, adjusted by the
variance-covariance matrix. The one-sample Hotelling's T-squared test statistic is used with a
significance level of 0.0500.

This report gives the power for each value of N and K. Notice that the power for K = 1 and N = 25 is 0.3379. This
is slightly different than the 0.3397 obtained by interpolation by Rencher.
Means Section
Means Section
Name Mean
Y1 1.8800
Y2 1.8800

This report shows the mean differences that were read in. When a Means Multiplier, K, is used, each value of K is
multiplied times each of these values.

Variance-Covariance Matrix Section

Variance-Covariance Matrix Section

Response Y1 Y2
Y1 7.5353 0.2938
Y2 0.2938 5.4111

601-7
NCSS, LLC. All Rights Reserved.
PASS Sample Size Software NCSS.com
Hotellings One-Sample T2

This report shows the variance-covariance matrix that was read in from the spreadsheet or generated by the
settings of on the Covariance tab. The standard deviations are given on the diagonal and the correlations are given
off the diagonal.

Chart Section

These charts show the relationship between power and N for each value of K.

601-8
NCSS, LLC. All Rights Reserved.