Maths & Statistics Home Which Statistical Test?  

If you know what you need to do, just select it from the menu above but if you don't know what statistical test to use, the menu below this text is a very simple and basic expandable menu that should tell which statistical test you should use depending on the type of data that you have. Just select the appropriate answer to each question to go to the next level.

Under "Tools", you will find "Terminology": a list of words that are specific to maths and statistics with their definitions which are applicable to this section. If you hover over some of the key words in the following menu, a box should appear with the definition or explanation (it may take a few seconds to appear).

To start, you must decide if you are testing for a difference in the data or an association:

How many sets of data do you have?

What kind of data do you have? Chose paired if your data is matched or related but chose unpaired if it is independent or unmatched.

What kind of distribution does the data show? (chose Non-Normal if you have tried but can't tell)

Use the Related t-Test

Requirements:

1.Matched / Repeated data
2.Interval data
3.At least 15 pairs of data
4.Normally distributed

If the data is not normally distributed or you don't have at least 15 pairs of data, use the Wilcoxon Test.

Use the Wilcoxon Test

Requirements:

1.Interval data
2.Between 5 and 15 non-zero differences
3.Normal or non-normal distribution of data
4.Matched / Paired data

If you have 15 or more pairs of data that are normally distributed, use the Related t-Test instead.

What kind of distribution does the data show? (chose Non-Normal if you have tried but can't tell)

Use the Student t-Test

Requirements:

1.At least 15 samples in each set of data (In an ideal world... I would expect at least 10 for an Extended Essay but for Internal Assessment, a minimum of five will be expected with an understanding that this reduces the validity of the analysis: hint hint wink nudge for the Evaluation...)
2.Interval data
3.Unmatched / Independent data
4.Normal distribution

If data is not normally distributed, use the Mann Whitney U Test.

Use the Mann Whitney U Test

Requirements:

1.Unmatched / Independent data
2.Ordinal or Interval data
3.Minimum 7 values for each set of data
4.Unknown or data with a non-normal distribution

If you have at at least 15 sets of data that are normally distributed, use the Student t-Test.

 

 

Use the ANOVA Test

Requirements:

1.Three or more sets of data
2.At least 15 samples in each set of data
3.Interval data
4.Unmatched / Independent data
5.Normal distribution

At which level is your data?

Use Pearson's Correlation

Requirements:

1.Six pairs of data (Five is possible for Internal Assessment with an understanding that this reduces the validity of the analysis: 2.hint hint wink nudge for the Evaluation...)
3.Normal distribution
4.Interval data

If the data is not normally distributed, use the Spearman Rank Correlation.

Use the Spearman Rank Correlation

Requirements:

1.Ten pairs of data
2.Interval or Ordinal data
3.Normal or non-normal distribution

If the data is at the interval level and normally distributed, use Pearson's Correlation.

How many variables have been changed to get the data?

Use the One Column X2 Test

Requirements:

1.Categorical data
2.No expected value is less than 5

Use the Two Column X2 Test

Requirements:

1.Categorical data
2.No expected value is less than 5