Mann-Whitney U-Test

Why do I need to do this test?

You need to perform this test if you are looking for a difference between one data variable at two locations or times.  It will identify whether the mean of one set of data is significantly different (higher or lower) to the mean of another set of the same data collected from a different place or time.

What do I do first?

Plot your data on maps or graphs.  This technique can be used with data collected using random sampling or systematic sampling.  You can use this method to statistically analyse isoline maps.  You can read about the test in Lenon and Cleves, Fieldwork Techniques and Projects in Geography pages 149-150. 

How do I perform the test?

You can do the test by hand using the instructions in Lenon and Cleves, or the class handout, or you can cheat a little and use an online Mann-Whitney U-Test calculator.

You may still have to put your result into a table (Lenon and Cleves pages 164-165)  to establish statistical significance.

What does the result mean?

The result will tell you if there is a statistical difference in the mean of the data from your two places or times.

The closer your value is to 0 the better your result.

Low values suggest a more reliable result, rather than a larger difference.

How do I use this result in the exam?

You can use your result in a question concerning analysis, conclusions or evaluation.

You may need to say why you did the test, what the outcome was (your Mann-Whitney U-test number), and what this tells you about the data.

The result of your test might support your hypothesis, or it might not - you need to say this.

You need to suggest the other factors that might have influenced the variations in your data, and you need to understand why - to do this you need a good grasp of the under-pinning theory.

Sample Paragraph for Conclusions.

The study aimed to investigate the difference between light intensity in a deciduous woodland and a coniferous woodland in the Wyre Forest, near Kidderminster.  Using an isoline map, a difference was identified which agreed with the hypothesis that light intensity would be higher in the deciduous woodland.  Following this, a Mann-Whitney U-test value of 0.126 was obtained.  This supports the hypothesis and makes the overall conclusion to this aim more reliable since this result is statistically significant at the 90% level.  However, since the Mann-Whitney U-Test value is more than zero, there must still be other factors that control light intensity other than vegetation type.  Slope aspect plays a part, whereby south-facing slopes in the northern hemisphere recieve greater insolation than north-facing ones. Since the data was collected across a valley, this factor may have accounted for lower light intensity results collected from the north-facing slope.  The tree density of the woodland may also have an impact on light intensity since the coniferous woodland is a plantation, and trees are planted at a higher density to increas profitability.  The branches are in close proximity and reduce the amount of light that can penetrate.