PSYC2017- CHAPTER 8

• Bivariate Correlation


  The association that involves two variables
• Bivariate


  Bi (two) - variate (variables)
• Correlation


  Linear association between two variables
• Correlation coefficient (Pearson's r)


  Used to denote the correlation between two variables
• Magnitude and direction of correlation coefficient

  • Magnitude of the absolute values reflect the strength of the correlation
  • Sign of the correlation coefficient reflect the direction of the correlation
• Continuous Variable (interval)


  A variable with an "infinite" number of possible values
• Categorical (Nominal) Variable


  A variable with non-numerical categories as possible values
• Describing association with categorical variables


  1. Use t-test to describe the difference in a continuous variable between groups of a categorical variable

  2. T-Test results can only be transformed into a Pearson's correlation coefficient
• People with siblings


  Tend to be more responsible than single children
• Number of pets owned

  Directly correlated with empathy for others
• Age


  Associated with higher earning
• Parents


  More likely to drive a mini-van than people without children
• Correlational study


  A study where the variables (2 or more) of interest are measured, not manipulated
• Construct validity


  • How well was each of the variables measured?
  • Does the measure have good reliability (Cronbach's alpha)?
  • What type of reliability is this a reflection of?
• Rule of thumb: Look for Cronbach's alpha values above 0.70
• Statistical validity


  • To what degree do the statistical results support the claims made in the study?
• Statistically significant


  The probability that the observed findings are observed due to chance is less than 5% (p < 0.05)
• Effect size


  • Describes the strength of an association
  • Larger effect size allows for more accurate prediction
• The effect size statistical does NOT tell you if the effect is significant
• Factors affecting statistical validity


  • Outliers could bias the observed relationship between two variables
  • Restriction of range: when certain range of a variable is artificially excluded from the data, the observed correlation is attenuated (lowered)
• What to do with outliers

  1. Determine if outliers are due to user/experimenter error or a legitimate data point

  2. Outliers can be removed if its due to an error

  3. Data should be analyzed with and without outlier to examine differences in outcomes

  4. Researcher should be transparent
• When range is restricted, the observed correlation is attenuated (lowered)
• Internal validity


  • Does the study allow us to make a strong inference regarding the nature and direction of the associations?
• Money


  Often associated with happiness and well-being, but does not mean money can buy happiness
• Correlation does not imply causality
• External validity


  • The degree to which the results from a study generalizes to other samples and situations
• Moderating variable


  A variable that affects the strength and/or direction of the relationship between two other variables
• Moderation effects can be used to test for external validity while interaction effects can not