Stata 14.2 se

Descriptive and correlational analyses were conducted using STATA 14.2/SE (StataCorp, 2015 ). Autoregressive cross-lagged coefficients were estimated using structural equation modeling with Mplus 7.4 (Muthén & Muthén, 1998–2017 ). The variables included in our final models contained 13.9% missing data, on average, ranging from 11.2% to 35.9%. Results from the Little’s (1988) missing completely at random test indicated that the data were not missing at random (χ² = 292.3, degrees of freedom = 246, p = 0.023). Full information maximum likelihood was used to address missing data, given that it is a less biased and efficient practice than ad hoc missing data methods (Newman, 2014 (link)). As shown in Figure 1, autoregressive paths were estimated for both parenting stress and children’s behavioral problems across the four time points at the ages 3, 5, 9, and 15. Cross-lagged pathways and transactional effects were estimated between parenting stress and children’s behavioral problems over time.

Taking into account a cluster effect and an alpha level of 0.05, a sample size of 1530 residents was initially calculated to achieve an absolute precision of estimate of ± 1% with a confidence level of 95% and an expected prevalence of 12% for MRSA, 6% for ESBLE and 0.5% for both VRE and CPE.
Data were analysed using STATA 14.2 SE (StataCorp LP, Texas, USA). Median and interquartile ranges (IQRs) were calculated for continuous variables. Prevalence of MDRO carriage was calculated for each MDRO (number of residents with MDRO per 100 screened residents). The calculated prevalence rates were weighted, taking into account the number of residents actually tested in each NH, compared to the theoretical number of residents to test in each NH in the study. Poisson distribution was used to calculate the 95% confidence intervals (95%CI).
In order to explore risk factors of MDRO carriage, odds ratios (OR) and 95%CI were calculated using logistic regression analysis. All predictors with p-value < 0.10 in univariate analysis were included in multiple logistic regression models with stepwise backward elimination of the least significant variable until all remaining variables had p-value < 0.05.

Descriptive data analysis (e.g., prevalence of exposures) was conducted using Stata 14.2 S/E (StataCorp, College Station, TX). Latent class models were estimated using MPlus version 7.4 (Múthen & Múthen, Los Angeles, CA).

A sample size of 1000 was sufficient to detect an effect size (Cohen f²) of 0.02 for a multivariable model comprising of 10 covariates to explain group variance and allow for separate models for trained and untrained HCWs in quantitative analysis. Data was captured with Epi Data 3.0 and analysed using STATA 14.2 SE (StataCorp.2012. College Station, TX: StataCorp LP). Data was weighted, adjusted and computed within 95% Confidence Intervals (CIs) and at significance of p < 0.05. Summary statistics are presented as percentages (%), mean scores (x̅) and standard deviations (SD) for all continuous variables. Chi-square tests for categorical variables, independent Student’s T tests for continuous variables and One-way Analysis of Variance (ANOVA) assessed the differences at 18, 24 and 36 months. Post-hoc Tukey tests were implemented to identify the variables involved. A score ≥ 50% was adopted for retention of knowledge. We investigated the relationship between retention of knowledge and covariates by logistic regression. To investigate the factors responsible for retention of knowledge we carried out multivariable logistic regression analysis by backward selection in an unadjusted model with all significant (p < 0.05) factors in the univariate analysis and adjusted model for region. Model assumptions were checked and variables were checked for collinearity and correlation.