2.4. Statistical Analysis

Data analyses were performed in R statistical software (open source (GNU license) statistical environment available with libraries (stat, pROC, plyr, ggplot2) at www.r-project.org (accessed on 26 February 2021)), version 3.6.0, using the following software: stat, plyr, ggplot2, pROC. The results were presented as the mean (standard deviation) for continuous variables and count (frequency) for discrete data. A p-value < 0.05 was regarded as statistically significant. Discrete data were compared for the groups with Pearson’s χ² test, with appropriate modifications (i.e., Yates’s correction, Fisher’s exact test or V² test). For ultrasonographic features differentiating pulmonary fibrosis from heart failure, independently and in complex models, a receiver operating characteristic (ROC) curve was plotted, and the area under the curve (AUC) was calculated, determining whether it differed statistically by 0.5 with the application of the DeLong test. AUCs for differentiating parameters and predictive models were compared with the DeLong test. For quasi continuous variables (e.g., a total number of intercoastal spaces containing consolidations) optimal cut-off points were determined with two methods (“closest topleft” and Youden). For all diagnostic parameters, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the likelihood ratio for a positive LR(+) and negative result LR(−) were calculated. For statistically significant models, logistic regression was performed, calculating the odds ratio for a positive result of the tested model and a respective Akaike information criterion (AIC) value. AIC allows for the comparison of different models, where the lower its value, the better a given model is adjusted to the experimental data.