On-line statistics

What type of data do you have?

Some statistical tests are sensitive to the type of data you have and it is necessary to identify which data you have before you choose a test.

Firstly all measurements fall into one of two overriding categories continuous or discontinuous.

Continuous - Measurements of length are continuous whereby fractions can be included. If you have rounded up to whole figures you can consider your data to be continuous.

Discontinuous - Generally counts of things (frequencies) whereby a fraction of an individual is impossible (although you may only have part of that individual). For example, if an insect with only its head and thorax falls into a pitfall trap it is a count of one, although it is only part of the individual. Remember that values derived from such data should be rounded to whole individuals.

Data then falls into several other categories in order of complexity starting with the simplest.

Nominal - according to categories i.e. species of plant or colour of eyes. These data are most often associated with the Chi² and contingency table. But they are not excluded from other tests.

Ordinal - categories but in ascending or descending ranks.

Interval - describes a nominal data set in which the units are the same size throughout the scale i.e. the difference between 21 and 27 is the same as between 1 and 7. Temperature is a good example. However, they do not have a zero value so cannot be said to be x amount of times bigger.

Ratio - This is the highest level of data complexity and can include all the previous categories. Such scales have zeros and are continuous. Linear dimensions are a good example.

In addition to the above there are other categories that the data may also fall into. It is not necessary to apply these to statistical tests unless stated.

Qualitative - do your data describe a quality rather than a measurement? i.e. species, male/female or colour describe qualities and normally the data that follow are counts for these categories.

Quantitative - measurements of a variable normally indicate a quantitative variable. Length of femur or weight of gonads are quantitative values.

Derived variables - these are data that have not been measured directly but have been calculated from measurements. The most common derived variables are proportions (ratios and percentages). Although you will not be restricted by the test you use proportions often do not follow normal distributions. Such data can be transformed to normality by using "arcsine". Transform all values in your sample by square rooting and finding the angle (degree or radian) for the sine of these values.

---

Descriptive Stats

Diversity Indices

Comparisons

Correlations

Regression