An Evaluation of the Use of Statistical Procedures in Soil Science

Abstract

Experimental statistical procedures used in almost all scientific papers are fundamental for clearer interpretation of the results of experiments conducted in agrarian sciences. However, incorrect use of these procedures can lead the researcher to incorrect or incomplete conclusions. Therefore, the aim of this study was to evaluate the characteristics of the experiments and quality of the use of statistical procedures in soil science in order to promote better use of statistical procedures. For that purpose, 200 articles, published between 2010 and 2014, involving only experimentation and studies by sampling in the soil areas of fertility, chemistry, physics, biology, use and management were randomly selected. A questionnaire containing 28 questions was used to assess the characteristics of the experiments, the statistical procedures used, and the quality of selection and use of these procedures. Most of the articles evaluated presented data from studies conducted under field conditions and 27 % of all papers involved studies by sampling. Most studies did not mention testing to verify normality and homoscedasticity, and most used the Tukey test for mean comparisons. Among studies with a factorial structure of the treatments, many had ignored this structure, and data were compared assuming the absence of factorial structure, or the decomposition of interaction was performed without showing or mentioning the significance of the interaction. Almost none of the papers that had split-block factorial designs considered the factorial structure, or they considered it as a split-plot design. Among the articles that performed regression analysis, only a few of them tested non-polynomial fit models, and none reported verification of the lack of fit in the regressions. The articles evaluated thus reflected poor generalization and, in some cases, wrong generalization in experimental design and selection of procedures for statistical analysis.

[No Title Available]

“…The first step in developing SPEED Stat was to clearly define the target audience experimental profile. This profile was utilized to define a list of useful analytical procedures based on the literature, especially the data obtained by Tavares et al (2016).…”

“…(uni-factorial), 86 treatments (bi-factorials with up to 10x8 + 6 treatments) or 486 treatments (tri-factorials with up to 10x8x6 + 6 treatments). The maximum number of replicates supported per treatment is only eight, which, according to Tavares et al (2016), is sufficient for the vast majority of experiments in soil science. Although a little restricted, this capacity is sufficient for most of the tests of value for cultivation and use, such as those necessary for registering cultivars.…”

“…Parametric statistical analysis provides the basic support for interpreting and making decisions about experimental data. Unfortunately, the statistical procedures employed to analyze experimental data in the agrarian sciences are frequently misused, as evidenced by Lúcio et al (2003), Kramer et al (2016), Tavares et al (2016), and Possatto Júnior et al (2019). Such misuse may be associated with a lack of statistical science knowledge; however, low-quality statistical analysis also can be associated with statistical software, which is difficult to use.…”

“…For example, Tavares et al (2016) verified that in the soil sciences, the renowned SAS (Statistical Analysis System, SAS ® Institute Inc., Cary, NC, USA) and R (R Project for Statistical Computing) systems are not among the most commonly used software, suggesting that they are not intuitive. In part, the difficulties using a statistical application are linked to the vast number of procedures available in such software (lack of conciseness and specific focus, both typical of non-minimalist design).…”

Trends in Brazilian Congress of Soil Science publications: An analysis of the 1947-2017 period

“…There are some studies in the literature that have evaluated soil science publications in Brazil, considering general aspects (Garrido & Garrido 2007, Prado 2008, Trajano et al 2013 or thematic areas (Barreto et al 2009, Bressiani et al 2015, Mores et al 2016, Tavares et al 2016). However, there are few studies of this nature that have looked at CBCS publications, which have the potential to reveal important information about the development of soil science in Brazil, given their scientific and temporal coverage.…”