Unit 4: Correlation - Subjective Questions

Correlation refers to the statistical relationship between two variables. It indicates the direction and strength with which changes in one variable are associated with changes in another variable.\n\n- A positive correlation means that both variables tend to increase or decrease together.\n- A negative correlation means that one variable tends to increase as the other decreases.\n- A zero correlation indicates the absence of a linear relationship.\n\nCorrelation is important in psychological research because it helps researchers study relationships among variables such as intelligence, anxiety, academic achievement, stress, and memory. It is also useful for prediction and for examining relationships when experimental manipulation is not possible. However, correlation does not establish causation.

Correlation may be classified according to its direction as follows:\n\n- Positive correlation: Both variables move in the same direction. For example, study time and examination scores may show a positive correlation.\n- Negative correlation: The variables move in opposite directions. For example, stress and psychological well-being may show a negative correlation.\n- Zero correlation: There is no systematic linear relationship between the variables. For example, shoe size may have approximately zero correlation with memory ability in a group of adults.\n\nThe direction of correlation is represented by the sign of the correlation coefficient. A positive sign indicates a positive relationship, while a negative sign indicates an inverse relationship.

The three types of correlation differ as follows:\n\n| Type | Direction | Interpretation |\n|---|---|---|\n| Positive correlation | Variables change in the same direction | Higher values of one variable are associated with higher values of the other |\n| Negative correlation | Variables change in opposite directions | Higher values of one variable are associated with lower values of the other |\n| Zero correlation | No consistent direction | Changes in one variable are not systematically associated with changes in the other |\n\nFor example, intelligence and academic performance may be positively related, anxiety and academic performance may sometimes be negatively related, and birth month and reasoning ability may show little or no correlation. The direction alone does not indicate the strength of the relationship.

The strength of a correlation refers to the degree to which two variables are related. It is represented by the absolute value of the correlation coefficient, usually denoted by .\n\n- indicates no linear relationship.\n- Values close to indicate a strong positive relationship.\n- Values close to indicate a strong negative relationship.\n- Values close to indicate a weak linear relationship.\n\nFor example, indicates a strong positive correlation, whereas indicates a weak negative correlation. The sign indicates direction, while the absolute value indicates strength. A correlation coefficient cannot be greater than or less than .

Pearson's product-moment correlation coefficient is a statistical measure of the strength and direction of the linear relationship between two quantitative variables. It is represented by .\n\nThe formula is:\n\n\n\nImportant assumptions include:\n\n- Both variables should be measured at the interval or ratio level.\n- The relationship should be approximately linear.\n- The observations should be independent.\n- The variables should not contain serious outliers.\n- For significance testing, the variables are generally assumed to be approximately normally distributed.\n\nPearson's ranges from to .

Pearson's correlation coefficient is based on the covariance between two variables divided by the product of their standard deviations. Its conceptual form is:\n\n\n\nUsing deviation scores, let and . Then:\n\n\n\nSubstituting and gives:\n\n\n\nFor raw scores, the computational form is:\n\n\n\nThis formula standardizes covariance, making the coefficient independent of the measurement units.

The following steps are used to calculate Pearson's by the deviation-score method:\n\n1. Arrange paired scores for variables and .\n2. Calculate the means, and .\n3. Find the deviation scores, and .\n4. Calculate , , and for every pair of observations.\n5. Obtain , , and .\n6. Substitute the values in:\n\n\n\n7. Interpret the sign as the direction and the absolute value as the strength of the correlation.\n\nThe pairing of scores must be maintained throughout the calculation because each score must correspond to the correct score.

The coefficients can be interpreted as follows:\n\n- : This indicates a strong positive linear relationship. Higher scores on one variable tend to be associated with higher scores on the other variable.\n- : This indicates a moderately strong negative linear relationship. Higher scores on one variable tend to be associated with lower scores on the other variable.\n- : This indicates an extremely weak or negligible linear relationship. The variables show almost no linear association in the sample.\n\nThe correlation coefficient describes association, not causation. A strong correlation does not prove that one variable causes the other.

The coefficient of determination is obtained by squaring Pearson's correlation coefficient:\n\n\n\nIt represents the proportion of variance in one variable that is statistically associated with variance in the other variable in a linear relationship. For example, if , then:\n\n\n\nThus, 36% of the variance is explained or shared linearly, while the remaining 64% is not explained by this relationship. The coefficient of determination has no negative sign because squaring removes the direction. It should not be interpreted as proof that one variable causes the other.

Pearson's correlation coefficient has several limitations:\n\n- It measures only linear relationships. A nonlinear relationship may produce a low even when the variables are strongly related.\n- It is sensitive to outliers, which can substantially increase or decrease the coefficient.\n- Correlation does not establish causation. A third variable may account for the observed relationship.\n- Restricted variability can reduce the size of the correlation.\n- Measurement error can weaken the observed relationship.\n- Pearson's may be inappropriate for ordinal data or markedly non-normal data.\n\nResearchers should inspect scatterplots, consider the measurement level of the variables, and use an alternative such as Spearman's correlation when appropriate.

A simple correlation describes the relationship between two variables, such as anxiety and test performance. It is usually represented by .\n\nA multiple correlation describes the relationship between one criterion variable and a set of two or more predictor variables taken together. It is represented by . For example, academic achievement may be predicted jointly from intelligence, study time, and motivation.\n\nImportant distinctions include:\n\n- Simple correlation involves one predictor and one criterion variable.\n- Multiple correlation involves several predictors and one criterion variable.\n- Simple correlation may be positive or negative, whereas the multiple correlation coefficient is non-negative.\n- indicates the overall strength of prediction from the group of predictors.\n- indicates the proportion of criterion variance jointly explained by the predictors.

The multiple correlation coefficient, denoted by , is the correlation between the observed values of a criterion variable and the values predicted from a linear combination of predictor variables .\n\nIt may be expressed as:\n\n\n\nwhere represents the predicted value of . The value of ranges from to . A value close to indicates that the predictors jointly provide accurate prediction of the criterion, while a value close to indicates weak joint prediction.\n\nThe squared multiple correlation, , represents the proportion of variance in explained jointly by the predictors. It is important to distinguish joint prediction from the unique contribution of each predictor.

For a criterion variable and two predictors and , the squared multiple correlation coefficient is:\n\n\n\nHere, is the correlation between and , is the correlation between and , and is the correlation between the two predictors. The multiple correlation coefficient is obtained by taking the positive square root:\n\n\n\nThe formula shows that the value of depends not only on the correlations between each predictor and the criterion but also on the correlation between the predictors. Highly overlapping predictors may contribute less unique information.

Predictor intercorrelation refers to the relationship among the independent or predictor variables in a multiple correlation analysis.\n\n- If predictors are moderately related to the criterion but are also highly correlated with each other, they may provide overlapping information.\n- High intercorrelations can create multicollinearity, making it difficult to determine the unique contribution of each predictor.\n- Adding a predictor does not necessarily produce a large increase in if it duplicates information already provided by other predictors.\n- Predictors that are relatively independent may contribute more unique information and increase the multiple correlation more substantially.\n\nTherefore, researchers should examine the correlation matrix and interpret both the overall and the unique contribution of individual predictors carefully.

Pearson's and Spearman's correlations are both measures of association, but they differ in important ways:\n\n| Feature | Pearson's correlation | Spearman's rank-order correlation |\n|---|---|---|\n| Data | Interval or ratio data | Ordinal, ranked, or non-normal quantitative data |\n| Relationship | Linear relationship | Monotonic relationship |\n| Calculation | Uses raw scores and deviations | Uses ranks and rank differences |\n| Sensitivity | More affected by outliers | Generally less affected by extreme scores after ranking |\n| Symbol | | or |\n\nPearson's correlation is preferred when its assumptions are satisfied. Spearman's correlation is useful when data are ordinal, when assumptions of normality are doubtful, or when the relationship is monotonic but not necessarily linear.

Spearman's rank-order correlation is a nonparametric measure of the strength and direction of a monotonic relationship between two variables. It is represented by or .\n\nIt should be used when:\n\n- The variables are measured at the ordinal level.\n- Scores can be meaningfully ranked.\n- The assumptions required for Pearson's correlation are not satisfied.\n- The relationship is monotonic but not necessarily linear.\n- The data contain outliers that may unduly influence Pearson's coefficient.\n\nSpearman's method replaces raw scores with ranks and then assesses the association between the ranks. It ranges from to , with the sign indicating direction and the absolute value indicating strength.

When there are no tied ranks, Spearman's rank-order correlation is calculated using:\n\n\n\nwhere:\n\n- is the difference between the paired ranks, .\n- is the squared rank difference.\n- is the number of paired observations.\n\nProcedure:\n\n1. Rank the scores of both variables.\n2. Assign the same ranking direction to both variables.\n3. Calculate the difference between paired ranks.\n4. Square each difference and find .\n5. Substitute the values into the formula.\n6. Interpret the resulting coefficient.\n\nIf the rankings are identical, and . If the rankings are exactly opposite, .

The calculation can be performed as follows:\n\n1. List the paired observations for variables and .\n2. Rank the scores from lowest to highest or highest to lowest.\n3. Rank the scores using the same direction.\n4. If tied scores occur, assign them their average rank.\n5. Compute the rank difference, , for each pair.\n6. Square each difference and calculate .\n7. For data without ties, use:\n\n\n\n8. Interpret the sign and magnitude of .\n\nFor tied ranks, a tie-adjusted procedure or the Pearson correlation computed on the ranks should be used.

Tied ranks occur when two or more observations have the same score. The usual procedure is to assign each tied observation the mean of the ranks that they would occupy. For example, if two observations occupy positions 3 and 4, each receives rank .\n\nAfter assigning average ranks, researchers may calculate Spearman's correlation as the Pearson correlation between the two sets of ranks:\n\n\n\nwhere and are the ranks of the two variables. For small samples or many ties, tie-correction procedures may be applied. Simply using the no-ties formula without considering ties can produce an inaccurate result.

A linear relationship occurs when changes in one variable are associated with proportional changes in another variable and the points can be approximated by a straight line. Pearson's correlation is designed primarily to measure this type of relationship.\n\nA monotonic relationship occurs when one variable generally increases as the other increases, or generally decreases as the other increases, although the rate of change need not be constant. The relationship may be curved but must maintain a consistent direction. Spearman's correlation measures monotonic association.\n\nFor example, a steadily increasing curved relationship may have a high Spearman correlation but a lower Pearson correlation. Thus, a low Pearson coefficient does not always mean that no relationship exists.