Starting with our pool of nine datasets (SLO, ISI, AMG, NOV, KAT, SHA, SAE, SIR, PHI), all datasets were first checked for uniqueness and then cross-checked against each other to retrieve a set of unique and independent datasets.
Completed questionnaires were first checked and coded.
The data were first checked for missing values.
For the narrative task, the assumptions were first checked.
The filled-in questionnaires were first checked and coded.
Predictions by a conventional slope stability analysis were first checked against a continuum-mechanics based numerical analysis.
Pre-selected potential items for use were first checked for possible outliers that may affect the goodness of fit of the model, using bivariate scatter plots and histograms.
Data residuals were first checked for normal distribution, investigating the skewness and kurtosis z-values as well as the Shapiro-Wilk test p value, before testing for homogeneity of variance.
The results of decomposition were first checked on similarity of recorded and reconstructed voltammograms followed by determination of individual concentrations by least square method (linear model fitting) and feed-forward artificial net consisting of two hidden layers with 7 and 3 neurons.
Loci were first checked for deviations from Hardy-Weinberg equilibrium in CERVUS 3.0 [31] (Table 1).
The results were first checked for their normal distribution by the F-test and then compared using Student's t-tests.
