What is a Friedman Test?
Download as PPTX, PDF8 likes6,201 views
The Friedman test is a non-parametric statistical test that is used as an alternative to repeated measures ANOVA when the data is ordinal or not normally distributed. It compares the median values across three or more related groups or time points. Unlike ANOVA, it does not assume normal distribution of the data or equal intervals between ranks. The Friedman test determines if there are statistically significant differences in the distributions of a dependent variable among multiple groups or time points. It is appropriate to use when the underlying data is on an ordinal scale or highly skewed.
What is a Friedman Test?
- 1. Friedman Test The non-parametric analogue to the repeated measures ANOVA is the Friedman test.
- 2. First let’s review what a repeated measures ANOVA is.
- 3. First let’s review what a repeated measures ANOVA. A repeated measures ANOVA is used to determine if there is a statistically significant difference between the results of the same measure with the same group across multiple occasions.
- 4. For example, Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 Player 2 Player 3 Player 4 Player 5 Player 6 Player 7
- 5. Here is the fictitious data: Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 Player 2 4 Player 3 3 Player 4 5 Player 5 4 Player 6 3 Player 7 4
- 6. Here is the fictitious data: Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 Player 2 4 2 Player 3 3 3 Player 4 5 2 Player 5 4 1 Player 6 3 2 Player 7 4 3
- 7. Here is the fictitious data: Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 2 Player 2 4 2 8 Player 3 3 3 9 Player 4 5 2 8 Player 5 4 1 7 Player 6 3 2 1 Player 7 4 3 8
- 8. Here is the fictitious data: Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 2 Player 2 4 2 8 Player 3 3 3 9 Player 4 5 2 8 Player 5 4 1 7 Player 6 3 2 1 Player 7 4 3 8 Mean 3.9 3.4 6.9
- 9. *Note 1: Remember that a Repeated Measures ANOVA does not pin point which pair is significantly different. It just let’s you know that there is a difference. Post hoc tests must be performed to determine which pair is different.
- 10. *Note 2: Because a pair-wise t-test is generally used to determine if there is a difference between two separate observation occasions, the Repeated Measures ANOVA is generally used with three or more occasions even though it could be used with just two.
- 11. The Friedman test is appropriate to use when the underlying measurement is on an ordinal scale or the distribution of a dependent variable is highly skewed. The Friedman test will estimate whether there are significant differences among distributions at multiple (more than two) observation periods.
- 12. The Friedman test is appropriate to use when the underlying measurement is on an ordinal scale or the distribution of a dependent variable is highly skewed. The Friedman test will estimate whether there are significant differences among distributions at multiple (more than two) observation periods. While the Repeated Measures ANOVA compares group means, the Friedman Test compares group medians. Remember that a Median is less resistant to outliers
- 13. So let’s go back to our pizza-eating football player example: Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 2 Player 2 4 2 8 Player 3 3 3 9 Player 4 5 2 8 Player 5 4 1 7 Player 6 3 2 1 Player 7 4 3 8 Mean 3.9 3.4 6.9
- 14. Notice that the highlighted numbers are outliers in each group which can skew the results for a Repeated Measures ANOVA. Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 2 Player 2 4 2 8 Player 3 3 3 9 Player 4 5 2 8 Player 5 4 1 7 Player 6 3 2 1 Player 7 4 3 8 Mean 3.9 3.4 6.9
- 15. Because the Median is less resistant to these outliers, notice how the Median shows more of a separation between these groups. The Median in this case may reflect reality better than the Mean. Number of Pizza Slices Eaten by a Group of Football Players Before During After Player 1 5 14 2 Player 2 4 2 8 Player 3 3 3 9 Player 4 5 2 8 Player 5 4 1 7 Player 6 3 2 1 Player 7 4 3 8 Mean 3.9 3.4 6.9 Median 4.0 2.0 8.0
- 16. Friedman test is also used when the data is rank ordered or ordinal (no equal intervals).
- 17. So let’s say that football players have to rate the pizza on a scale from 1-10 in terms of its tastiness. Our hypothesis is that the pizza tastiness scores will change based on the different tasting occasions.
- 18. So let’s say that football players have to rate the pizza on a scale from 1-10 in terms of its tastiness. Our hypothesis is that the pizza tastiness scores will change based on the different tasting occasions. Football Players Opinions on Degree of Tastiness of Pizza at a Certain Pizza Café on a Scale of 1-10 Before During After Player 1 5 10 1 Player 2 6 8 2 Player 3 5 9 1 Player 4 6 9 2 Player 5 6 7 1 Player 6 7 10 2 Player 7 6 9 3 Median 6.0 9.0 2.0
- 19. Once again the median is used here because on a scale of 1-10 there is not an agreed upon definition usually on how a 1 differs from a 2 compared to how a 7 differs from an 8. Rating pizza taste is more subjective in terms of how equal the intervals are compared to the number of pizza slices which is a more objective measure.
- 20. Once again the median is used here because on a scale of 1-10 there is not an agreed upon definition usually on how a 1 differs from a 2 compared to how a 7 differs from an 8. Rating pizza taste is more subjective in terms of how equal the intervals are compared to the number of pizza slices which is a more objective measure. That is why in this case we would use a Friedman Test