PSYC1040 Week 8

Question 1

Shared or ‘explained’ variance: the coefficient of determination

Answer 1

• generally, whenever we have an association between two variables, we can say one of them ‘explains variance’ in the other
• the amount of variance explained depends on the strength of the association
• in a linear correlation, the proportion of variance explained or shared is equal to r2
• TOTAL VARIANCE OF Y - calculated from deviations of each Y from the mean of Y

Question 2

Regression line

Answer 2

• a line drawn to minimise the vertical distance between itself and all points
• it represents the linear correlation of x and y
• a sloped line indicates that the correlation is not zero and same variation in y is explained by x

Question 3

partitioning (dividing) variance

Answer 3

1. total variance - based on the deviation of raw scores from the mean
2. explained variance - based on deviations of predicted scores from the mean (shared variance)
3. unexplained variance - based on deviations of raw scores from predicted scores
   - coefficient of determination (r2)

Question 4

Null hypothesis significant testing: assessing chance as alternative explanation ~

Answer 4

• alternative explanations may be raised by problems with the study design, quality of operations, sampling methods and soon
• one alternative explanation must always be dealt with in change
  ~ in particular; the possibility that a result was due to sampling error
• conventionally, the probability of the obtained result under sample error done must be LESS THAN 5% for sampling error to be discarded as a plausible explanation

Question 5

Null hypothesis significant testing: assessing chance as alternative explanation (continued) ~

Answer 5

• null hypothesis - due to chance
• alternative hypothesis - not due to chance
• two possible states of reality
• four possible outcomes
• THE NULL HYPOTHESIS IS TRUE - sampling error caused the result
• THE HYPOTHESIS IS FALSE - sampling error did not cause the result
• RETAIN THE NULL HYPOTHESIS - attribute the result to sampling error
• REJECT THE NULL HYPOTHESIS - attribute the result to something else - accept the alternative hypothesis