2) The type of statistical tests used is dependent on the experimental design. A review by Mischener 1997 outlined common experimental designs used in restoration ecology and described appropriate statistical tests for each.
3) There has been a push to incorporate more variables in experimental designs to better explain systems. There is also movement away from ANOVAs towards Generalized Linear Models or GLMs. These movements have likely impacted the experimental designs, statistics and even the citation rate of restoration studies.
4) The purpose of this review is to look at the statistical scope of restoration ecology of invaded grasslands. I will examine the type of statistical tests being over time, whether assumptions are tested, and how statistical tests and the experimental design of studies influence the citation rate and the impact of the journal
5) A systematic literature search was conducted and resulted in 340 articles. Each article was individually screened to exclude reviews and irrelevant articles resulting in 103 retained articles
6) ANOVAs are the most commonly used statistical test comprising more than half of all tests. T-tests, non-parametric tests, and tests classified as ‘other’ are used less often.
7)ANOVAs have consistently been the most common statistical test. After 2000, the diversity of tests increased to include the diversity of tests observed today
8) The majority of studies tested the assumptions of the statistical tests used in their analyses.
9) Articles that performed more statistical tests have a citation rate and are published in higher impact journals. Increasing the number of factors has a negative impact on both the citation rate impact factor of journals.
10/11)The sample size of studies is positively correlated with both the average citation rate of the article and the impact of the journal where the article was published.
12) ANOVAs are likely the most common test as it is appropriate for many study designs used in restoration ecology. However, one assumption that ecological data infrequently meets without transformation is the assumption of normality. Thus, I recommend a shift towards GLMs which are appropriate for many study designs in restoration ecology and can handle non-normally distributed data
13) Although the majority of studies did test the assumptions of the statistical tests used, there is still many studies that did not test the assumptions which calls into question the interpretations of their results. Statistical models are also often poorly described in terms of how they defined, and this needs to be improved upon in future studies
14) Study design and statistics have an influence on the citation rate of articles and the impact factor of the journal where these articles are published. In order to increase the citation rate and journal impact, the sample size and number of tests should be increased. The number of factors should not be increased at the expense of sample
4. Purpose: to describe the statistical scope of
studies restoring invaded grasslands
Most common statistical tests over time
Statistical reporting
Influence on citation rate and journal impact
5. Methods: systematic literature search and
PRISMA report
Search terms:
(restor*) AND (grass* OR
savanna) AND (non-native* OR
invas* OR invad* OR alien)
Refined to “ecology”
12. Discussion: a shift from ANOVAs to GLMs is
recommended
GLMs do not need a normally distributed response
GLMs are appropriate for many study designs
13. Discussion: statistical reporting needs
improvement
Assumptions must be tested
Statistical models need to be appropriately described
14. Discussion: statistics influences the citation
rate and journal impact factor
When possible sample size should be increased
Number of factors should not be increased at the
expense of sample size
15. Conclusion:
Shift from ANOVAs to GLMs
To increase confidence we need to improve reporting
Statistics and experimental design should be carefully
chosen
