Video Tutorials: Statistical Methods When Using Routine Data in Evaluation
Numerous data analysis methods are used in evaluation and research, ranging from bivariate to more advanced methods like time-series. There is also an increased use of routine data in evaluation and research. Yet, practitioners across the world are not always able to spare the time to take entire courses on data analysis—particularly during this time of COVID-19 when it is increasingly difficult to create in-person experiences.
These online data tutorials focus on two common methods identified in a literature review of methods used when evaluating routine data: interrupted time series and logistic regression; each of which answers different types of questions.
The tutorials pose example research questions then present possible answers, as well as step-by-step instructions on how to analyze and interpret the data using Stata. Sample data sets, .do files, and background materials are provided as supplemental files for each method.
Tutorial 1: Interrupted Time Series Analysis
Time series analysis is a statistical technique used for trend analysis or time series data. As the name implies, time is an important factor. Time is often the independent variable and can be monthly, weekly, daily, annual or even biennial. This method is useful for aggregate level analysis such as social statistics including divorce rates, crime rates and often useful in evaluating the impact of social or political policies. In some instances, time series has been used for forecasting where historical values and associated patterns have been used to predict future activity.
The following videos and supplemental files focus on the interrupted time series analysis.
Interrupted Time Series Video Tutorial: Part 1
Interrupted Time Series Video Tutorial: Part 2
Interrupted Time Series Video Tutorial: Part 3
Tutorial 2: Logistic Regression
Logistic regression, also referred to as binary logistic regression, is a specialized form of regression analysis used when the outcome variable is a nonmetric, binary/dichotomous variable. Often, researchers want to analyze whether an event has occurred or if a person falls within a target group. It has extensive use in healthcare and social sciences and is often used to develop prediction models that predict the probability of developing a disease outcome. Additionally, logistic regression is easy to use and produces robust estimates.
The following video and supplemental files focus on the logistic regression.