RELAN
The New approach of logically and causally based statistical analysis
Why use RELAN?
Why use RELAN instead of conventional methods? The RELAN Software enables:
- Gain insight into logical relations between a set of variables
- Analyse and statistically test verbally formulated questions and hypotheses
- Evaluate causal assumptions between a given set of variables
- Simulate regularities in a sample of data (including their statistical significance
- Conduct simple and transparent significance testing (e.g., Binomial tests, Likelihood-ratio test, Chi- square tests)
- RELAN analysis takes into account all possible relations between a defined set of dichotomous variables
- RELAN is a strictly multivariate oriented method which tests conditional, moderating, mediating and multicausal influences on criteria variables
- Provides more than fifty significance and effect statistics
- Relationships between variables that might dilute or falsify results - confounding effects - can be extracted from the tested hypothesis
- Null-Hypothesis testing is completed by an alpha correction procedure (Bonferroni correction method)
- Because of alternative hypothesis testing results can be evaluated by the power statistic ( = generalisation of results)
- The option of bivariate testing runs (logically) multivariate and includes seven logical functions (AND, OR, IF-THEN, …)
- The bivariate relations between the variables can be analysed graph theoretically (features of networks between variables and their relations
- Hypothetical relations can be simulated in an user-defined set of data (including testing the significance of the simulated relation)
Benefits of the RELAN Software
- The opportunity to test hypotheses logically allows a better statistical fitting of common speech questions (ecological validity)
- The (logically) multifunctional approach opens deeper insights into hidden relations between variables
- Simulation option shows how the data structure had to be if the hypothetical relation would be given ideally
- Simple and complex relations of confounding variables can be extracted from the hypothesis relation
- The option of implication analysis highlights the existence of one-sided perhaps causal relations
- Moreover, the option of causal analysis gives the opportunity to test causal assumptions more precisely based on time indices of variables
Key Features
RELAN is a non-parametric statistical method
- which processes propositional logical hypotheses,
- needs (or computes) dichotomous variables, and
- analyses simple and complex logical functions (AND, OR, IF-THEN, NAND, …) between variables.
RELAN can
- test the logical functions statistically or explore them by data mining procedures,
- eliminate confounding effects,
- evaluatedifferent chance models
- perform “real” causal analyses by taking into account the implication relation between the variables and the time relations between the variables, and
- determine the necessary sample size for a special logical hypothesis.
On top of that the concept of RELAN can be summarized by the following features:
- RELAN refers both to a specific research logic and to the statistical procedure of the computer program which can be downloaded from this page
- The type of used variables is dichotomous (two-valued) or can be dichotomized by the program
- There are no preconditions of variable distributions (except for “ergodicity” of the probabilities)
- Seven logical functions are evaluated (conjunction, disjunction, adjunction, bijunction, implication, NAND, NOR) and statistically tested or can be explored by procedures of data mining
- Chance model selection: The probabilities of the positive values of the variables can be defined apriori or will be computed aposteriori depending on the data
- If for the used set of variables time indices are available then the probability of existing causal relations can be evaluated by the program
Methodological Overview
A relation analysis refers to four possible ways of computations:
- Boolean analysis of hypotheses uses propositional logic and logical decomposition of the hypotheses.
- Statistical analysis of the distribution of frequencies in the “mintermes” (all possible combinations of variable values).
- Causal analysis of statistical results by using the causal indices of variables, in which the onset and the duration of corresponding events are defined.
- Simulation of relations for the purpose of testing the logical structure of specific hypothesis (e.g., to find contradictions, tautologies, subrelations and so on) and testing its significance for a specific sample size.
Fact Sheet
This fact sheet contains a description of the main purposes, features, coding languages, program flow, file names etc. of the RELAN software.
Statistics
RELAN computes more than fifty statistics most of them well-known in statistical practice.
