Notes on Normality, Erroneous Data, and Data Screening

• Understanding key concepts related to data screening is crucial for ensuring the integrity and validity of statistical analyses.

• Ensure that each student collects data from three participants.
• Deadline for data collection is set for next Friday (28th).

• After this lecture, you should be able to:
• Screen and check your data effectively.
• Identify erroneous data and outliers, understanding the next steps upon discovery.
• Interpret histograms and assess normality tests.
• Comprehend the definition of a normal distribution and its implications.
• Take appropriate actions when data is not normally distributed.
• Differentiate between parametric and non-parametric tests.

• Data from studies can consist of:
• Demographic data
• Variables of interest, such as:
  • Scores on questionnaires
  • Reaction times
  • Physiological data

• Check for:
• Erroneous data
• Normal distribution of data

• Erroneous data are values that could misrepresent the population or are incorrect due to methodological errors or issues in data collection. Examples include misreported scores due to misunderstanding questions, or outlier responses that are implausible given the context.

• Check demographic data for consistency
• Time check to assess if participants took a reasonable duration to complete surveys
• Verify responses for expected variability and ranges

• Age, condition, and other relevant variables must be evaluated for outliers or erroneous entries. Example: responses outside the expected age range (e.g., age < 18 in an adult sample).

• Utilize JAMOVI for summarizing descriptive statistics and visualizing data through histograms, box plots, and frequency distributions.
• Example output: Age demographics with mean = 29.3, median = 22.0, standard deviation = 13.1, and age range of 17-61.

• Remove any participant data that does not meet foundational criteria and document the reasons for removal. E.g., if a participant reports being under 18 in a study focused on adults.

• Analyze how much time it took participants to complete surveys to identify outliers (e.g., too short of a response time).
• Generate descriptives and frequencies to understand average completion times (e.g., mean = 157 seconds).

Importance of Checking Responses

• Ensure responses fall within expected ranges based on validated scales. For instance, using the Comfort from Companion Animals scale, responses should yield scores that reflect the range from 1 (strongly disagree) to 4 (strongly agree).

• Checking Individual Items

  • Always remember to analyze summed scores for subscales to account for potential errors in computation.

• Judiciously evaluate outliers in context and document actions taken with precision. Use descriptive statistics and visual tools to verify unusual score patterns.

• Effective data screening is essential to ensure robust conclusions. Focus on identifying:
• Missing data
• Data out of expected ranges
• Correct computation of derived scores.

• In quantitative research, statistical tests determine if the scores are statistically different or related. The choice of test (parametric vs. non-parametric) depends on:
• Design
• Data type
• Assumptions met by the data.

• Visualize distributions using histograms to assess normality, noting skewness and kurtosis as key indicators.
• Use tests such as Shapiro-Wilk to statistically determine normal distribution (p-value > .05 suggests normality).

• Ensure the application of tests is based on adherence to assumptions of the data (i.e., normality, homogeneity of variance, and data scale).

• Workshop activities will allow practice in checking normality, identifying erroneous data, and drafting methods sections for research reports.
• Remember to distribute and gather data for your surveys to meet the collection deadline next Friday.