Home / Summaries / Class notes - Ecological Methods I / group-post-hoc-anova

Artikel Z - Tutorial

Q: How do you report the results of an Anova test?

F1,18=2.205, p>0.05) the F-value the two degrees of freedom of the between and within sum of squares in subscript

Q: When are you allowed to use an Anova?

Data follow normal distribution Data has equal variances

Q: How does the Bonferroni-correctionare work and what are the advantages and disadvatages?

By calculating a new critical p-value by dividing p=0.05 by the number of comparisons that you want to make (in this case we would have 3 comparisons: fast-slow, fast-intermediate, slow-intermediate) so the critical p-value becomes 0.05 / 3 = 0.016. Advantages : easy to apply over a wide range of applications provides sooner significant results if the number of comparisons is low Disadvantages : increase in Type-II error when the number of comparisons increase (test becomes too conservative , critical p-value is too small)

Q: Why is the LSD test a liberal post-hoc test?

Because it does not correct for the number of multiple comparisons you carry out. For this reason, it has relatively great power , and hence finds more significant differences than the HSD test

Q: What is a downside of the LSD test?

It results in a relatively large Type 1 error (false positives). Therefore, it is recommended to use the LSD test only with well-justified hypotheses (i.e. the hypothesis is strongly expected to be confirmed based on earlier research, for example the hypothesis of a positive association between smoking and lung cancer) or with very obvious hypotheses (for example the hypothesis of a positive association between rainfall and vegetation biomass such as illustrated here).

Q: What is a trick you could do to carry out a Post-hoc test after an Anova with unequal samples sizes?

To randomly select the same number of samples from each group (= a subset), and carry out the p ost-hoc test for these equal sample sizes . The group size depends on the group with the lowest sample size (in this case 7), and you could thus take 7 random samples from each of the other groups.

6 important questions on Artikel Z - Tutorial

How do you report the results of an Anova test?

F1,18=2.205, p>0.05)

the F-value
the two degrees of freedom of the between and within sum of squares in subscript

When are you allowed to use an Anova?

Data follow normal distribution
Data has equal variances

How does the Bonferroni-correctionare work and what are the advantages and disadvatages?

By calculating a new critical p-value by dividing p=0.05 by the number of comparisons that you want to make (in this case we would have 3 comparisons: fast-slow, fast-intermediate, slow-intermediate) so the critical p-value becomes 0.05 / 3 = 0.016.

Advantages:

easy to apply over a wide range of applications
provides sooner significant results if the number of comparisons is low

Disadvantages:

increase in Type-II error when the number of comparisons increase (test becomes too conservative, critical p-value is too small)

Why is the LSD test a liberal post-hoc test?

Because it does not correct for the number of multiple comparisons you carry out.

For this reason, it has relatively great power, and hence finds more significant differences than the HSD test

What is a downside of the LSD test?

It results in a relatively large Type 1 error (false positives). Therefore, it is recommended to use the LSD test only with well-justified hypotheses (i.e. the hypothesis is strongly expected to be confirmed based on earlier research, for example the hypothesis of a positive association between smoking and lung cancer) or with very obvious hypotheses (for example the hypothesis of a positive association between rainfall and vegetation biomass such as illustrated here).

What is a trick you could do to carry out a Post-hoc test after an Anova with unequal samples sizes?

To randomly select the same number of samples from each group (= a subset), and carry out the post-hoc test for these equal sample sizes.

The group size depends on the group with the lowest sample size (in this case 7), and you could thus take 7 random samples from each of the other groups.

The question on the page originate from the summary of the following study material:

Ecological Methods I

View summary