About
usConservationEducationWildlifeHabitatsMapsSearchResourcesSponsorsAn Ecological Sampling Exercise
Processing & Interpreting Raw DataExercise Answers 5 and 6PreviousSection 1 (autumn)Section
2 (spring)Section
3 (autumn)Section 1 (autumn)Section
2 (spring)Section
3 (autumn)Section
4 (spring)Section
4 (spring)A test is required which will test the difference
between two sets of data. The t-Test is one such statistical test. However, this test
should not be used on small sets of data, which are not normally distributed. Most of the
Bluebell data is not normally distributed, but is skewed (see below). Strip
1Strip 2Strip 3Strip 4(undivided)TopBottomTopBottomTopBottom%
cover ClassNo
of Samples0
- 103131341011
- 202452121
- 3013112131
- 401141
- 5023151
- 6021161
- 7011171
- 8011181
- 901191
- 100Another such test is the
Mann-Whitney U Test. This would be more suitable as it is not dependent on data being
normally distributed (it is 'distribution-free'). This test is also suitable for small
sets of data ( 6 to 20 measurements). It is not suitable for sample sizes of greater than
20. (For the workings of a Mann-Whitney U Test carried out on the data click here)Median values for the Bluebell % cover data
are calculated below for comparison with the means.Data sets ranked in order,
with comparisons of mean and median valuesAutumn
CutSpring cutAutumn CutSpring CutQuadrat No.Strip
1Strip 2Strip 3Strip 4(undivided)TopBottomTopBottomTopBottom158802589031010100420101005201513063015201745152418501525197020271109020402112045412204561320456142549815255515162560301725606018406775198070208580Mean3218.147.928.324.41.5Overall34.52538.113Median22.517.54522.511.51.0Overall252042.51.5Because the data is not normally distributed,
outlying values which are unusually high or low have a disproportionate influence on the
mean. This can therefore disguise real differences between the data sets. In this case,
the median is probably a better figure to represent the mid point of individual data sets.Answers
Contents