Example Paired Ttest PDF
Example Paired Ttest PDF
It is thought that the glycaemic index (GI) of food is an indicator of how sustaining
or satisfying it is and may influence appetite.
A sample of children were provided with a breakfast of low GI foods on one day
and high GI foods on another. The two breakfasts contained the same quantities
of carbohydrate, fat and protein. On each day a buffet lunch was provided and
the number of calories eaten at lunchtime were recorded.
On the first day the children ate a low GI breakfast and on the second day a high
GI breakfast. The objective is to determine whether the kind of breakfast eaten
has an effect on mean calorie intake.
The table below summarises the data for children in the sample. This shows that
for the children in the experiment, mean calorie intake at lunchtime was higher
after the high GI breakfast. A hypothesis test is needed to determine whether
these results mean that we can predict that there would be differences in mean
calorie intake for other children who ate low and high GI breakfasts.
Paired Samples Statistics
Std. Error
Mean N Std. Deviation Mean
kcal after low gi breakfast 607.9697 33 182.82445 31.82565
kcal after high gi breakfast 771.6061 33 222.10779 38.66400
The hypotheses to test are:
Null hypothesis (H0): Mean calorie intake is the same following low or high GI
breakfast
Alternative: (H1): Mean calorie intake depends on whether a low or high GI
breakfast was eaten
A paired ttest is appropriate because:
(1) two related samples are being compared. The samples are related
because the same group of children tested both breakfasts
(2) the dependent variable, that is, the calorie intake, is a continuous variable
The next table shows the result of the paired ttest – this allows us to draw
conclusions about the population of children from which the sample was drawn.
The key outputs are the test statistic, the degrees of freedom and the Pvalue.
Paired Samples Test
Paired Differences
95% Confidence
Interval of the
Std. Error Difference
Mean Std. Deviation Mean Lower Upper t df Sig. (2tailed)
Pair kcal after low gi
1 breakfast kcal after 163.636 186.40762 32.44940 229.734 97.53910 5.043 32 .000
high gi breakfast
SPSS calculates the Pvalue from the test statistic and the degrees of freedom
and reports it to 3 decimal places. The actual value of P is 0.0000175. This is so
small that when reported to three decimal places, it is shown as 0.000 which we
report as P < 0.001. Because the Pvalue is less than 0.05, we have statistically
significant evidence that mean calorie intake at lunchtime depends on whether a
child’s breakfast had a low or high GI.
Although our conclusion is based wholly on the Pvalue, the convention is to
report the value of t and the df as well.
The hypothesis test allows you to say:
The kind of food eaten at breakfast has a statistically significant effect on mean
calorie consumption ( t= 5.043, df= 32, P < 0.001).
The confidence interval allows you to say:
We are 95% confident that eating a low GI breakfast reduces children’s mean
calorie consumption at lunchtime by between 98 and 230 kcal (to the nearest
whole number).
Assumptions underlying the test
The paired ttest is based on the assumption that if you calculate the difference in
calorie intake following low and high GI breakfasts for each child, the results are
Normally distributed. If there are major problems then the results of the test can
be unreliable. This can be checked by plotting the differences between intakes
following low and high GI breakfasts on a PP plot. This is a graph that is
specially designed to detect departures from the Normal distribution (which
appear as deviations from the 45 degree line). If the data does not support the
Normality assumption, then the problem can be solved either by using an
alternative test, such as the Wilcoxon test or by transforming the data.