Ecological data vary naturally, so a difference or trend in results could be due to chance. Statistical tests tell us how likely it is that a result occurred by chance.1. State a null hypothesis (H0): there is no significant difference between the samples, or no significant correlation between the variables. For example: there is no significant difference in the mean shell height of limpets on exposed and sheltered shores.
2. State the alternative hypothesis (H1): there is a significant difference or correlation.
3. Choose the correct test:
- correlation between two variables: Spearman's rank
- difference between two means of normally distributed data: Student's t-test
- difference between observed and expected frequencies: chi-squared (χ2)
4. Calculate the test statistic and the degrees of freedom.
5. Compare it with the critical value from a table, at the 5% significance level (p = 0.05).Using p = 0.05 means we accept a 5% probability that the result is due to chance. If the calculated value is greater than or equal to the critical value, the probability of the result being due to chance is 5% or less, so we reject the null hypothesis: the difference or correlation is significant.If the calculated value is less than the critical value, we accept the null hypothesis: any difference is not significant and may be due to chance.Remember that a significant correlation does not prove causation. Another variable may be causing both changes.
2. State the alternative hypothesis (H1): there is a significant difference or correlation.
3. Choose the correct test:
- correlation between two variables: Spearman's rank
- difference between two means of normally distributed data: Student's t-test
- difference between observed and expected frequencies: chi-squared (χ2)
4. Calculate the test statistic and the degrees of freedom.
5. Compare it with the critical value from a table, at the 5% significance level (p = 0.05).Using p = 0.05 means we accept a 5% probability that the result is due to chance. If the calculated value is greater than or equal to the critical value, the probability of the result being due to chance is 5% or less, so we reject the null hypothesis: the difference or correlation is significant.If the calculated value is less than the critical value, we accept the null hypothesis: any difference is not significant and may be due to chance.Remember that a significant correlation does not prove causation. Another variable may be causing both changes.