Using SurveyMars: Likert, Stratified Sampling & Ordinal

Introduction

Survey research is a powerful tool for gathering data, but its effectiveness depends heavily on proper methodology. This article will guide you through using SurveyMars to create professional surveys that incorporate three critical elements: Likert scales for measurement, stratified random sampling for representative data collection, and proper handling of ordinal scale data. When combined effectively, these components can significantly improve the quality of your survey results.

Understanding the Key Concepts

Likert Scale: The Foundation of Attitude Measurement

The Likert scale, developed by Rensis Likert in 1932, is one of the most reliable methods for measuring attitudes, opinions, or perceptions. Typically structured as a 5point or 7point scale ranging from "Strongly Agree" to "Strongly Disagree," it transforms subjective responses into quantifiable data.

Common Likert scale examples include:

Customer satisfaction surveys ("Very satisfied" to "Very dissatisfied")

Employee engagement surveys ("Strongly agree" to "Strongly disagree")

Product feedback questionnaires ("Extremely likely" to "Extremely unlikely" to recommend)

Stratified Random Sample: Ensuring Representative Data

Stratified random sampling involves dividing your population into homogeneous subgroups (strata) and then randomly sampling from each stratum. This approach:

Improves representation of key subgroups

Increases statistical efficiency

Allows for separate analysis of different strata

For example, when surveying employee satisfaction, you might stratify by department, job level, or tenure to ensure all groups are adequately represented.

Ordinal Scale: Understanding Your Data Type

Ordinal scale data has a clear ordering but uncertain intervals between values. Likert scale responses are ordinal  we know "Agree" is more positive than "Neutral," but we can't quantify exactly how much more. This distinction is crucial for proper statistical analysis.

Implementing These Concepts in SurveyMars

Step 1: Designing Your Likert Scale Questions

1. Log in to SurveyMars and create a new survey

2. Add a "Matrix/Rating" question type for Likert scales

3. Define your scale points (typically 5 or 7)

4. Phrase statements carefully to avoid bias

5. Balance positive and negative statements to reduce acquiescence bias

Pro Tip: Use an even number of points (e.g., 6) if you want to force respondents away from neutral options.

Step 2: Setting Up Stratified Sampling

1. Identify your strata variables (demographics, usage patterns, etc.)

2. In SurveyMars:

    Use screening questions to identify strata membership

    Set quotas for each stratum under "Survey Settings"

    Enable random sampling within each quota group

3. Determine sample size for each stratum (proportional or equal allocation)

Example: For a customer satisfaction survey, you might stratify by:

Age groups (1825, 2635, etc.)

Subscription tier (Basic, Premium)

Frequency of use (Weekly, Monthly)

Step 3: Configuring Ordinal Data Collection

1. In SurveyMars question settings:

    Select "Ordinal" as the measurement level

    Ensure response options are logically ordered

    Avoid assigning numerical values that imply interval properties

2. For analysis:

    Use nonparametric tests (MannWhitney U, KruskalWallis)

    Report medians and interquartile ranges rather than means

    Consider visualization with ordered bar charts

Best Practices for Implementation

1. Pilot Testing: Run a smallscale test to:

    Verify strata definitions

    Check Likert scale interpretation

    Identify unclear questions

2. Response Options:

    Label all scale points (not just endpoints)

    Maintain consistent direction (positive always on same side)

    Consider visual analog scales for enhanced precision

3. Sampling Considerations:

    Ensure strata are mutually exclusive and collectively exhaustive

    Balance between too many strata (small cell sizes) and too few (loss of precision)

    Document sampling methodology for transparency

Analyzing Your Data in SurveyMars

SurveyMars provides several tools for working with your stratified Likert scale data:

1. Descriptive Statistics by Stratum:

    Frequency distributions

    Median and mode calculations

    Visual comparisons across groups

2. Comparative Analysis:

    Nonparametric tests for ordinal data

    Stratumspecific vs. overall results

    Trend analysis across time (for longitudinal studies)

3. Advanced Options:

    Weighting adjustments for disproportionate sampling

    Reliability analysis (e.g., Cronbach's alpha for multiitem scales)

    Factor analysis for underlying dimensions

Common Pitfalls to Avoid

1. Treating Ordinal Data as Interval:

    Avoid calculating means of Likert scales

    Don't assume equal distances between points

    Use appropriate statistical tests

2. Sampling Errors:

    Inadequate stratum representation

    Failure to properly define strata boundaries

    Ignoring important stratification variables

3. Survey Design Issues:

    Doublebarreled Likert items

    Unbalanced response scales

    Leading or ambiguous questions

Conclusion

By combining Likert scale measurement, stratified random sampling, and proper handling of ordinal data in SurveyMars, you can create surveys that yield meaningful, representative, and statistically sound results. Remember that each component serves a specific purpose: the Likert scale captures attitudes, stratification ensures proper representation, and ordinal scale recognition guides appropriate analysis.

SurveyMars provides the tools to implement this sophisticated methodology without requiring advanced statistical knowledge. Following these guidelines will help you avoid common survey pitfalls and produce research findings that truly reflect the population you're studying. Whether you're measuring customer satisfaction, employee engagement, or public opinion, this integrated approach will significantly enhance the quality and reliability of your survey data.