Stata se version

Descriptive analyses included proportions, mean (SD) values, cross-tabulations, and data visualizations. We calculated the prevalence of multimorbidity in an age stratification model, varied by sex, race/ethnicity, and SES. Multimorbidity patterns compared against age and mental health multimorbidities were statistically different across the Chinese and Indian races/ethnicities (but not patients who were Malayan or of other races/ethnicities). Hence, post hoc tests were used to conduct pairwise comparisons for Chinese and Indian patients with multimorbidities. Differences in the prevalence of multimorbidity between different variable groups (eg, the prevalence of multimorbidity in female vs male patients) were measured using the χ² test of independence. We used logistic regression to examine the associations between mental health diseases and age, sex, and SES, while adjusting for the number of physical diseases. We identified the top 10 most prevalent chronic diseases, then assessed their co-occurrence with one another on the distribution of health care visits and costs in 2016.
Data cleaning and preparation were conducted on Python, version 2.7 (Python Software Foundation); statistical analyses were conducted using Stata/SE, version 14.0 (StataCorp). In 1-tailed and 2-tailed tests, P < .05 was considered statistically significant.

Statistical analyses and graphics were made using RStudio software, version 1.0.153 (https://rstudio.com/: accessed 19 August 2020) with R software, version 3.5.1 (http://www.r-project.org: accessed 19 August 2020) and Stata/SE, version 12.0 (StataCorp LLC, College Station, TX, USA). To compare physiological indices, bacterial relative abundance, alpha-diversities, and the level of bacterial metabolites, a Wilcoxon rank-sum test was used to compare two groups. Pairwise Wilcoxon rank-sum with Bonferroni or Holm adjustment were used to compare more than two groups, except for the comparison of NOO among the different BMI groups in which a Welch’s t-test was used. Regression and correlation analysis of bacterial abundance and other indices were calculated by the lm function in R for normally distributed independent variables, or Spearman’s rank correlation in Stata for non-normally distributed variables. Validation of the established linear model was performed using the gvlma function in R. For the linear regression analysis, regression of microbiome or host physiological indices onto PCA ordination was performed with the ordisurf function from the vegan package in R.

Statistical analysis was conducted at the outbreak level. Each risk factor was compared to each of the 3 severity measures using simple linear regression. For model selection, backward elimination and stepwise regression were performed using Akaike information criteria (AIC), and P values were used to identify significant independent risk factors. Multivariate Gaussian regression analysis was conducted for attack rate, and multivariate negative binomial analysis was used for the number of deaths and outbreak duration. Generalized estimating equations (GEE) were added as an extension to the models to account for clustering.^{14 (link)}
Goodness of fit was assessed by analyzing residual plots and deviance residuals. All models included age and sex to control for confounding. A P value of .05 was considered statistically significant. Statistical analyses were performed with SAS Studio version 3.8 software,¹⁵
R Studio version 4.1.0 software,¹⁶
and Stata/SE version 17.0 software.¹⁷