Statistical tools and packages for data collection, management, and analysis - A brief guide for health and biomedical researchers

Abstract

Previous articles in this series have looked at various aspects of planning, designing, conducting and interpreting biomedical research. In this article, we offer an overview of some tools and resources available to health and biomedical researchers, to help them with their research.

INTRODUCTION

The quality of a research study depends largely on the quality of its data. Moreover, the quality of a dataset is determined by its characteristics such as validity, accuracy, completeness, and consistency.[1,2] Efforts toward collecting good-quality data start right at the stage of developing a study protocol. To achieve high quality of data, a study must have an appropriate design with adequate sample size, and an efficient data collection mechanism with inherent data quality checks. Research study designs and principles of sample size calculation for different types of research studies have been discussed in previous articles.[3,4,5,6,7,8] In this article, we will provide an overview of some free or low-cost resources available for various research activities such as power and sample size calculation, randomization, data capture, data analysis, and visual representation of data. This is not meant to be a comprehensive overview since there are a very large number of resources, all of which cannot be covered.

POWER AND SAMPLE SIZE CALCULATIONS

Methods for sample size estimation are based on complex statistical theories spanning the concepts of probability, hypothesis testing, and confidence intervals. As a result, more often than not, sample size calculation may act as a deterrent for medical and public health researchers while developing a research protocol. However, for most of the standard study designs and research outcome measures, the resultant formulae for obtaining the estimate of adequate sample size are well established and popularly known in the literature. In addition, several web-based and software-based tools are available which provide platforms for the easy implementation of these formulae. Power of the test (or the concerned study) and sample size are functions of each other and knowledge of one is required to calculate the other. Accordingly, these tools provide options for calculating both sample size and power. A simple exploration on a search engine provides an overwhelming list of open access sample size and power calculators. However, one must be cautious in choosing a calculator. It is advisable that researchers use tools which are developed or hosted by a recognized institution/organization, and which preferably provide well-documented references for the methods. In addition, while these tools may act as a guide for researchers, it is always best to get the calculations verified by a qualified statistician.

RANDOM ALLOCATION

An important part of any clinical trial is the method used to allocate participants to the various treatment groups, which includes the process of generating a random number sequence (randomization) and the process used to conceal the sequence (allocation concealment). The ideal software used for the random allocation process should allow generation of study identity (to conceal participant identity), methods for variations of randomization (permuted block, stratified randomization, unequal allocation, etc.), and a mechanism to break the code if the protocol mandates it. Martin Bland has a web page (https://www-users.york.ac.uk/~mb55/guide/randsery.htm) which lists various resources (randomization software and randomization programs) for random allocation.

ELECTRONIC DATA CAPTURE AND MANAGEMENT

Electronic tools may be used either as an alternative to, or synchronously with paper data collection forms. A good electronic data capture system should be user-friendly, allow data validation, have an audit trail, export easily to a variety of statistical software, and have security features to protect data confidentiality. Additional provisions such as automatic creation of data dictionary or metadata and assignment of role-based access rights to research personnel improve the quality and ease of data collection and management.

DATA ANALYSIS

Data analysis may involve a wide range of statistical techniques depending on the study objectives, outcome measures, and the nature of the data. By and large, statistical data analysis can be broadly classified into six processes: data cleaning, descriptive analysis, estimation and hypothesis testing (statistical inference), correlation and regression analysis, nonlinear modeling, and multivariate analysis. Descriptive analysis helps to characterize the data through summary statistics and data visualization. It is fundamental toward developing a preliminary understanding of the data and identifying possible patterns and associations. It also helps in detecting data inconsistencies such as outliers, missing values, and violation of distributional assumptions, among others. While data cleaning techniques may be considered as precursors to data analysis, they are also employed for removing data inconsistencies identified through descriptive analysis. Some data collection and management tools, such as Research Electronic Data Capture, Epi Info, Census and Survey Processing System, etc., have built-in options for basic data analysis such as summary statistics, graphical visualization, and hypothesis tests.[9,10,11] Relatively more analysis options are available in spreadsheets such as Microsoft Excel (MS Excel) and Google sheets. MS Excel, which is a part of the Microsoft Office suite, is a popular spreadsheet application for collecting and storing data, and has an elaborate list of functions, calculations, pivots, and charts for analyzing the data. Availability of various add-ins in MS Excel, like the Analysis ToolPak, Solver, etc., make it possible to perform additional tasks such as hypothesis tests, correlation and linear regression analysis, basic time-series analysis, and optimization.

To carry out a more comprehensive data analysis, statistical software packages such as IBM-SPSS (IBM Corp., Armonk, New York, USA), Stata (StataCorp LLC, Texas, USA), R (R Foundation for Statistical Computing, Vienna, Austria), Python (Python Software Foundation, Lafayette Boulevard, Virginia, USA), SAS (SAS Institute Inc., North Carolina, USA), etc., can be used. Some of these packages, such as IBM-SPSS, Stata, and SAS have a menu-driven and user-friendly interface for nontechnical users which enables them to run even highly complex statistical analysis with just a few clicks.[12,13,14] Users who are comfortable in programming may prefer using commands and codes in Stata and SAS to widen their scope of analysis. Statistical packages such as SAS, IBM-SPSS, and Stata offer more advanced analysis options but are expensive and require some expertise.

For technical users, R and Python are among the most popularly used programming languages for statistical computing and graphics. Both are freely available software, with extensive coverage of statistical and computational methods through numerous packages.[15,16] R is specifically designed as an environment for statistical analysis and is an integrated suite of software facilities for data manipulation, calculation, and graphical display.[15]

Many other statistical software and tools are available as well. The choice of software for analysis should be based upon the level of analysis to be performed and the technical capacity of the user. In addition, researchers must ensure credibility of the software or tool, especially if they are not well known in academic circles. Authenticity of developers, validation of software and tools, and information on scientific peer-review or associated publications are some of the aspects to focus on before using any such software. Ideally, statistical analysis software should offer a wide range of statistical techniques, handle multiple types of data file formats, allow hassle free import and export of data files, have adequate options for graphs and visualizations, provide output in tangible formats, and preferably allow tracing and saving of command flow.

For a quick reference, a list of some easily accessible web-based resources and software tools with their features are provided in Table 1.

Table 1.

Some easily accessible web-based resources and software tools with their features

Resource	Features
PS: Power and sample size calculation	Free software
https://biostat.app.vumc.org/wiki/Main/PowerSampleSize	Allows calculations for studies with continuous, dichotomous or time-to-event outcomes
https://cqsclinical.app.vumc.org/ps	Has downloadable versions for iOS, Windows and Linux operating systems
	In addition, there is a web-based program
	Allows sample size calculation, power calculation, and detectable alternative hypothesis for a given sample size and power
REDCap	Free for consortium members
https://projectredcap.org	Server-based
	Create data entry forms and research databases
	Exports to most statistical software tools
	Allows multiple simultaneous projects and simultaneous access from multiple sites
	Mobile-based applications to increase functionality
	Facilitates electronic consent
	Audit trail
	Multiple language options
MS Excel	Part of Microsoft Office suite
	Spreadsheet for creating database and formatting
	Data files are compatible with all data analysis software and programming languages
	Has many built-in formulae for data analysis
	Provides option for user-defined formulae
	A wide range of visualization options through graphs
	Add-in features such as “Analysis ToolPak”
	Prompt-based AI co-pilot option available in recent versions in Microsoft 365
Google Sheets	Similar to Excel in many ways
	Differences are
	Needs internet connectivity to activate all features
	Allows collaboration - sharing of databases in real-time
	Has revision history - earlier versions can be accessed
Sealed Envelope	Randomization software
https://www.sealedenvelope.com/	Has a free option for first 50 randomizations
	Also has an online database application (Red Pill) for EDC and electronic patient reported outcomes
	Offers basic power and sample size calculation for trials with binary and continuous outcomes
MedCalc	A statistical software package
https://www.medcalc.org/	Requires purchase of license
	Trial version is free
	Data management options with integrated spreadsheet
	Includes more than 220 statistical tests, procedures and graphs, with ROC curve analysis, method comparison and quality control tools
Epi Info	Developed by Centers for Disease Control and Prevention
https://www.cdc.gov/epiinfo/index.html	Free resource: Can be used to create forms, collect data, and perform epidemiologic data analysis and visualization (graphs and maps)
	Used in epidemiology/public health
	Mobile version for use with tablets or smartphones to conduct epidemiologic studies in the field
NVivo	Used for the analysis of unstructured text, for example, interviews or focus groups
https://lumivero.com/products/nvivo/	Helps in transcribing and coding data and sorting into thematic areas
	Used in qualitative and mixed-methods research
	Trial version is free
CSPro	Free public domain software package for entering, editing, tabulating, and disseminating census and survey data
https://www.census.gov/data/software/cspro.html	Developed and supported by the U.S. Census Bureau and ICF Macro
	Supports data collection on android devices (phones and tablets)
	CSEntry Android App works in collaboration with the desktop version of of CSPro
	Supports smart data transfer from Android or Windows devices to a server running CSWeb
	Also contains a sophisticated programming language to create highly customized applications like quality checks
	Has freely available extensive learning resources (https://www.census.gov/programs-surveys/international-programs/events/training.html)
ODK	An open-source software for creating forms and collecting data
https://getodk.org/	Users can self-host and self-support it for free but requires technical expertise
	ODK Cloud is a paid version. It is the same ODK software, but fully hosted, managed
	Powerful data collection forms can be built with options for photos, GPS locations, skip logic, calculations, external datasets, multiple languages, and more
	Works both online and offline through mobile app and web app
	Offline data are automatically synced once internet is connected
	Option to connect with applications such as MS Excel, R, Python, or Power-BI to create real time dashboards

REDCap=Research Electronic Data Capture, EDC=Electronic data capture, ROC=Receiver operating characteristic, CSPro=Census and Survey Processing System, ODK=Open Data Kit, MS Excel=Microsoft Excel, AI=Artificial intelligence, PS=Power and Sample Size

CONCLUSION

There are several free and paid statistical software tools available to perform the tasks of sample size and power calculation, randomization, data collection, data management, and data analysis. In general, criteria for choosing a tool should include cost considerations, credibility of the tool, ease of use as per the technical capacity of the user, range of functions, and features such as audit trail and validation. Availability of such a wide range of options enables the researchers in the field of health and clinical research to effectively plan and execute their studies. However, researchers must ensure appropriateness of methods before implementing them through these tools and packages.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.