On-line statistics

Categories covered by this web site

Descriptive Statistics This must be your next stop!

Before we analyse data we need to know a few things about the data sets (samples) we have obtained. Firstly we need to know about the frequency pattern that the data follow (distribution) to indicate whether a parametric or non-parametric test must be used. Secondly we need to know how close the sample is to the population (see also Overriding limitations of statistics section).

Diversity Indices

Communities (species and relative abundance) require their own form of analysis. This section covers methods of calculating diversity indices and how to compare between two or more samples. Projects that have gathered the types of animal or plants and possibly the numbers of each type as in pit-fall trapping or quadrat sampling of plant species can use this form of test.

Comparisons

Characteristics of a species or of a response of a species can be different between two locations because of some environmental effect (rainfall), or some dietary factor or pollutant. The comparative tests can be used to analyse differences between two or more samples.

Correlations/Relationships/Trends

If two variables rise and fall with each other or as one rises the other falls they are said to be correlated either positively (as in the first case) or negatively (as in the second). The fecundity (egginess) of an animal may respond positively (go up) to increasing abundance of an essential food. Conversely, your bank balance will respond negatively (go down) as the number of evenings you spend in the pub increases. These are examples of relationships between two variables (fecundity/food availability and money in the bank/evenings out).

Predictions

Taking the above correlation examples a step further, if we know that an insect's reproductive output is going to be positively affected by increasing availability of a food plant can we estimate how many eggs an individual will lay for a given amount of this plant? By using regression we can. We can also suggest how close our predictions are likely to be to the real value.