Template for Statistical Analysis

It is often noticed that novice researcher finds it challenging to summarize various analyses carried out in his or her research. Statistical jargon required may be overwhelming for any researcher who lacks statistical expertise. Hence, a generic template for summarizing statistical tests and analyses carried out in a given research study is mentioned below.Data were explored for any outliers, typing errors and missing values. All quantitative variables were estimated using measures of central location (mean, median and mode) and measures of dispersion (standard deviation and standard error). Their 95% Confidence intervals were also calculated. Qualitative or categorical variables were described as frequencies and proportions. Normality of data was checked using graphics (histograms, box and whisker plots, Q-Q plots) and statistically by measures of skewness and kurtosis. Data were also presented graphically with box & whisker plots, histograms (with superimposed normal curves), bar diagrams (clustered as well as stacked) and with plotted pie charts as well. Chi-square test was used to find out any statistical association between categorical variables. Independent T-test and Mann Whitney U test were used to compare various quantitative variables between two groups. Relative risks were calculated for the various outcomes; say for example death, intubation, retinopathy, peptic ulcer perforation, pancreatitis etc. for two groups. General linear model analysis with Repeated Measure ANOVA was used to assess the trend for values serially measured during follow-up in two study groups. The mean values of age, and other quantitative variables were compared among 3(or >) study groups using one way ANOVA tests. Pearson’s correlation coefficients were also calculated between different quantitative variables. For serially measured values, repeated measure ANOVA was used to assess trend in change of serial values and interaction of trend in two study groups. Univariate odds ratios and their 95% C.I. were also calculated to identify predictors for the outcome of interest. Multivariate analysis was further carried out to find significant predictors for the outcome of interest after adjusting for various confounding variables. “Time to Event” data was analyzed using Kaplan Meier Survival Analysis and time to outcome, say for example, “time to cancer relapse” or “time to relapse to smoking ” after treatment for cancer or smoking addiction was compared between two groups using log rank sum test. All tests were two-tailed and p-value less than 0.05 was taken as significant.