Experimental design
In this tool, a one-way layout of Anovas has been employed. This means that the experimental design should be of the type where the outcome of a single continuous variable is compared between different groups. The tool cannot be used to analyze experiments where two or more variables vary together.
Tip: You can create a new column using the Concatenate function (or '&') of the Insert Calculated Column tool (Insert > Calculated Column...) if you want to analyze two or more variables together.
Distribution of data
The Anova and Linear regression comparisons assume the following:
The data is approximately normally distributed.
The variances of the separate groups, or the variances of the errors in the case of linear regression, are approximately equal.
If the data do not fulfill these conditions, the Anova and Linear Regression comparisons may produce unreliable results. In this case, it may be more valid to use a Kruskal-Wallis or Spearman R comparison instead.
Note: If more than one test is performed at the same time, then it is more likely that there will be at least one p-value less than 0.05 than in the case where only one test is performed. A guideline of when to reject the hypothesis is then "Reject the hypothesis if the p-value is less than 0.05 divided by the number of tests". This is called the Bonferroni method for multiple tests.
Back to Overview of Data Relationships theory