B io Factsheet

April 2001 Number 85

The Paired t-Test & When To Use It

The t-test is widely used within project work. This Factsheet will tell you when and how to use the paired t-test. Questions using the
t-test may occur on exam papers, but you will not have to remember the formula - it will be given to you.

Hypotheses and 1 or 2-tailed tests.

The t-test is used to compare two means (the mean is the normal "average"). As with all other statistical tests, this involves choosing between two hypotheses.
For the t-test, the null hypothesis (H0) is always that the two means are equal - for example "mean length of leaves at the top of the tree = mean length
of leaves at the bottom of the tree". It can never be that the means are different, or that one is bigger/smaller than the other. You always start out assuming
that the null hypothesis is true, and only change your mind if the evidence is good enough.

The other hypothesis - alternative hypothesis (H1) - can take one of three forms (Table 1); the precise form of it tells you whether you are carrying out
a one-tailed or a two-tailed test. A one-tailed test means that you are only considering one alternative - for example that the leaves are bigger at the top of
the tree. A two-tailed test means you are considering either alternative - that the leaves could be either bigger or smaller at the top of the tree.
Table 1. Alternative hypotheses for the t-test
Alternative Hypothesis Example 1-tailed or 2-tailed Use it if you are expecting....
mean 1 > mean 2 mean leaf length at top > mean leaf length at bottom 1-tailed leaves to be longer at the top
mean 1 < mean 2 mean leaf length at top < mean leaf length at bottom 1 tailed leaves to be shorter at the top
mean 1 mean 1 mean leaf length at top mean leaf length at bottom 2-tailed not sure - could be either!

Paired and Unpaired Tests

The t-test may be paired or unpaired. The calculations involved are When can the t-test be used?
different in each case. This Factsheet concentrates on the method for the You can only use the t-test when the population your observations come
(commoner) paired test; a later Factsheet will deal with the unpaired test. from is normally distributed. That means that if you were to collect the
same data from every member of the population and draw a graph of your
Paired tests are used when the data occur in "natural" pairs. Examples results, you would get a symmetrical, bell-shaped graph (Fig 1).
include:
the same organism reacting to two different stimuli.
Fig 1. Normal distribution curve
comparisons of plant growth between points in the same position on
different sides of a hedge.
No. of individuals with

measurements made on identical twins.

that measurement

Note that the pairs must be matched in a natural way - you cannot just decide
to pair results up.

The idea of pairing is to increase the chance that any differences really are
due to what you are investigating - e.g. which side of the hedge the point is
- rather than due to other variations - e.g. soil quality.

If a paired test is not applicable, then an unpaired test must be used. Measurement
Table 2 shows some examples of investigations requiring paired and What sorts of things are normally distributed?
unpaired tests. Normally distributed variables must be continuous - this means that they
can take any value (like lengths), rather than only set values (like numbers
Table 2. t-test investigations of organisms - you can't have 2.85 spiders!)
Investigation Paired or Unpaired?
Generally speaking, anything that is measured will be normally
Comparison of plant growth on two Paired distributed. This includes length, width, height, weight and velocity.
sides of a hedgerow
Effect of different soil types on the Anything that is counted (like numbers of organisms) or calculated (like
yield of a crop plant Unpaired
a diversity index) will not be normally distributed. In these cases, you
Effect of pollution on vegetation Unpaired should use the Mann-Whitney U-test instead.

Sample size
Exam Hint: - If you are unsure, it is better to use an unpaired test. You cannot use the t-test unless you have at least 6 measurements in each
Even if a paired test would have been suitable, it is not wrong to use an
sample for the paired test.
unpaired one (although it is harder to get a significant result). But
using a paired test where it is not appropriate is definitely wrong and
will not gain you marks! However, you will find it much easier to get a significant result if you have
larger numbers than this - aim for at least 10 in each sample.

1
The paired t-test Bio Factsheet

Worked Example - paired t-test

Eight adults who normally ate a "junk-food" diet agreed to follow a "healthy" diet for two months. The following data gives their weights in kilograms
immediately before and immediately after the two-month period.

Person A B C D E F G H
"Before" weight 76.4 85.9 50.6 62.4 90.5 55.7 78.3 67.4
"After" weight 74.5 80.1 50.8 62.3 79.4 55.7 73.2 65.0

Step 1: Write down the hypotheses, deciding which form Since we would expect a healthy diet to reduce weight compared to a junk-food
of the alternative hypothesis you want to (use and diet, we have:
so whether it is 1-tailed or 2-tailed). H 0: Mean weight "before" = Mean weight "after"
H 1: Mean weight "before" > Mean weight "after" So it is a 1-tailed test

Step 2: Decide whether it is paired or unpaired This is paired, because we are considering the same people "before" and "after"

Step 3: For a paired test, work out "1st value - 2nd value" Person A B C D E F G H
for each individual "Before - "After" 1.9 5.8 -0.2 0.1 11.1 0 5.1 2.4
(76.4 - 74.5)

Step 4: Work out the mean of these differences by adding 8 = 3.275

(1.9 + 5.8 + -0.2 + 0.1 + 11.1 + 0 + 5.1 + 2.4)
them up and dividing by how many there are.

Step 5: Calculate the standard deviation s, using x2 =1.92 + 5.82 + (-0.2)2 + 0.12 + 11.12 + 02 + 5.12 + 2.42 = 192.28
x
2 n=8
s= x2 Exam Hint: - When you are
n
s = 192.28 3.2752 = 3.648 working out x 2 , you must
8 square each value then add
where x2 means square each difference, then add up them - you cannot add then
n = no. of values square afterwards.
x = mean of the differences
= add up

Step 6: Work out the test statistic, using

x n 1 t-test = 3.275(8-1) = 2.375
t-test = 3.648
s

Exam
ExamHint
Hint:- -Do
Donot
notworry
worryabout
about
Step 7: Work out the degrees of freedom, using degrees of freedom = 8 - 1 = 7
what
whatthis
thismeans,
means,but butremember
remember
degrees of freedom = n - 1 totouse
usenn--11

Step 8: Get a t-test table and look up the value for the We are carrying out a 1-tailed test; Table 3. t-table
significance level (usually 5%), the degrees of freedom we will use a 5% significance level.
Significance level
and whether it is a 1-tailed or 2-tailed test. We need to use 7 degrees of freedom. d f 0.1 0.05 0.025 0.01 0.005
5 1.476 2.015 2.571 3.365 4.032
So our tables value is 1.895 6 1.440 1.943 2.447 3.143 3.707
7 1.415 1.895 2.365 2.998 3.499
8 1.397 1.860 2.306 2.896 3.355
9 1.383 1.833 2.262 2.821 3.250
10 1.372 1.812 2.228 2.764 3.169
11 1.363 1.796 2.201 2.718 3.106
12 1.356 1.782 2.179 2.681 3.055

Step 9: Make a decision - if your t-value is bigger than the Our test-value (2.539) is bigger than the tables value.
one from the tables, you can reject the null So we reject the null hypothesis.
hypothesis. Otherwise you have to accept it. So the mean weight before is significantly greater than the mean weight after.

2
The paired t-test Bio Factsheet

Practice Question Answer

The plant known as Indian Mustard (Brassica juncea) will accumulate gold (a) (i)
from gold salts in the soil in which it grows. It can be used to extract gold gold accumulated/
from impoverished mine wastes which still contain residual gold which is plant
mg gold Kg-1 dry mass
uneconomical to extract by traditional methods. Addition of ammonium x=b-a x2
thiocyanate to the mine waste growth substrate is thought to enhance the no with
accumulation of gold by the plants. An investigation was carried out to test thiocyanate(a) thiocyanate(b)
whether this was so. The results of the investigation are shown in the table
below.
1 1.01 0.99 -0.02 0.0004
gold accumulated/ 2 1.09 1.09 0 0.0000
plant
mg gold Kg-1 dry mass 3 0.98 1.25 0.27 0.0729
x=b-a x2
4 0.71 1.34 0.63 0.3969
no with
5 1.15 1.22 0.07 0.0049
thiocyanate(a) thiocyanate(b)
6 1.21 1.18 -0.03 0.0009
7 1.18 1.23 0.05 0.0025
1 1.01 0.99 8 0.89 1.34 0.45 0.2025
2 1.09 1.09 9 1.26 1.21 -0.05 0.0025
3 0.98 1.25 10 1.07 0.97 -0.1 0.0100
4 0.71 1.34 11 1.30 1.18 -0.12 0.0144
5 1.15 1.22 12 1.17 1.23 0.06 0.0036
6 1.21 1.18
7 1.18 1.23
8 0.89 1.34 x values;
9 1.26 1.21 x = 1.21;
10 1.07 0.97 x = 1.21 = 0.101; 3
11 1.30 1.18 12
12 1.17 1.23
(ii) x2 values;
The null hypothesis was proposed that: x2 = 1.464 ; 2

the mean mass of gold = mean mass of gold (iii) (x)2 = 0.010(2) ; 1
accumulated without thiocyanate accumulated with thiocyanate
(iv) s2 = 1.464 - 0.010 = 0.112 ;
The alternate hypothesis considered was: 12
mean mass of gold mean mass of gold s = 0.335 ; 2

accumulated without thiocyanate accumulated with thiocyanate
(v) t = 0.10112 1 ; = 1.00 ; (allow 0.999) 2
(a) Carry out a t-test to test the validity of the null hypothesis. 0.335

The formula for this t-test is: (n 1)

t = x (b) (i) accept the null hypothesis;
s calculated value is less than the critical value (at 0.05 probability
level/at all probability levels shown); 2
where x is the mean of the weight differences, n is the number of
readings in each group and s is found from the formula: (ii) use same strain/seed batch/similar Brassica seedlings;
mine waste material should have identical gold contents to begin
s2 = x2 - x2 with;
n use a standard concentration of thiocyanate;
(i) Calculate the value of x. 3 grow under similar conditions of temperature/humidity/water
(ii) Calculate the value of x2. Show your working. 2 availability; max 2
(iii) Calculate the value of (x)2. Show your working. 2 Total 14
(iv) Calculate the value of s. Show your working. 2
(v) Calculate the value of t. Show your working. 2

(b) At 11 degrees of freedom the critical values for t at various probabilities

are shown in the table.

Probability 0.1 0.05 0.02 0.01 0.001

Critical value 1.80 2.20 2.72 3.11 4.44

Acknowledgments: This Factsheet was researched and written by Cath Brown.
(i) Does the calculated value for t enable you to accept or reject the Curriculum Press, Unit 305B The Big Peg, 120 Vyse Street, Birmingham B18 6NF
null hypothesis? Explain your answer. 2 Bio Factsheets may be copied free of charge by teaching staff or students,
(ii) Suggest two precautions which should have been taken to ensure that provided that their school is a registered subscriber. No part of these Factsheets
the investigation yielded valid results. 2 may be reproduced, stored in a retrieval system, or transmitted, in any other
form or by any other means, without the prior permission of the publisher. ISSN
Total 14
1351-5136