Statistical decision tree

Anne-Maïmiti DulaurentLéa Kervroëdan

TO DO LIST

To import your data into R,

A "Stats" folder, and "Data" and "TD" subfolders you will create

In the "Data" folder the data as CSV you will save

The Excel file with the TD data in the folder "TD" you will put

In R, the "setwd" function you will use

The data, in relation to the desired statistical tests, you will arrange

With the "read.csv" function the data you will import

Univariate statistics

Sheet 1: Univariate statistics for 1 qualitative INDEPENDANT variable ("factor")

Normality test Shapiro-Wilk

p < 0.05

p > 0.05

-> Non normal distribution -> Non parametric tests family

-> Normal distribution -> Parametric tests family

2 Modalities

+ than 2 modalities

+ than 2 modalities

2 modalities

Paired

Paired

Paired

Unpaired

Unpaired

Paired

Unpaired

Unpaired

t-test

t-test

Wilcoxon test

Friedman test

Mann-Whitney test

Repeated measures ANOVA

+ Wilcoxon post-hoc test

+ post-hoc Tukey test

Kruskall-Wallis test

One-way ANOVA

+ post-hoc Tukey test

+ Mann-Whitney or Dunn post-hoc test

Sheet 2: Univariate statistics for 1 quantitative INDEPENDANT variable

Normality test Shapiro-Wilk

p < 0.05

p > 0.05

-> Non normal distribution -> Non parametric tests family

-> Normal distribution -> Parametric tests family

Pearson correlation test

Spearman correlation test

p < 0.05 and Corr coeff > 0.6

p < 0.05 and Corr coeff > 0.6

Non apparié

Simple linear regression

GLM (Generalised Linear Model

Multivariate statistics

Sheet 3: Multivariate statistics for several qualitative independant variables ("factors")

Effect on a quantitative dependant variable

Normality test Shapiro-Wilk

Effect of several (less than 5) qualitatives independant variables

Effect of 1 independant qualitative variable et 1 independant quantitative variable

Multifactorial ANOVA

PERMANOVA

ANCOVA

If normal distribution

If non normal distribution

explanatory methods with p-value

SHEET 4 : Multivariate statistics for several quantitative independant variables

Effect on a quantitative dependant variable

Normality test Shapiro-Wilk

p < 0.05

p > 0.05

Normal distribution

Non normal distribution

Multiple regression multiple

GLM (Generalised Linear Model

explanatory methods with p-value

SHEET 5 : Multivariate statistics for many qualitative or quantitative independant variables

Not linearly correlated variables

Transformation of linearly correlated variables into decorrelated variables

Automatic classification of individuals into a number of classes

Special case in ecology, adapted to the presence of a gradient of biological variables. Applies to living communities.

Quantitatives variables

HACHierarchical Ascending Classification

PCAPrincipal Component Analysis

DCA Detrended correspondence analysis

Contingency table

CACorrespondence Analysis

NMDS Non-metric Multi-Dimentional Scaling

Qualitative variables

MCA Multiple Correspondence Analysis

explanatory methods without p-value