Regression Flashcards
What are continuous independent variables sometimes termed as
Predictor variables
What does a significant effect in regression indicate
Predictive relationships (A predicts B) NOT Causal relationships (A affected B)
Does Regression deal with Continuous or Categorical IVs?
Continuous
Things a regression analysis can predict
Does a predictor significantly predict the DV
What is the strength and direction of the predictive relationship
How much variance in the DV is explained by the model
What regression equation best summarises the relationship
What additional things a multiple regression analysis can predict
Does a predictor significantly predict the DV when controlling for other predictors
Is the overall model significant
How much variance in the DV is explained by each predictor
What analysis options are there for the following study?
Q - does organisational identification predict commitment to an organisation
H - stronger organisational identification will predict stronger commitment (ie OA will positively predict C)
M - questionnaire measures of both variables
Linear regression
What does this linear regression output tell us
In this study about organisational identification predicting commitment - what is the regression equation
The first table tell us the significance of the overall model (it is significant)
It also explains the variance in the DV - r2 and adjusted r2
The second table tells us the significance of the predictor (in the organisational ID row)
It also gives us the regression coefficients (0.62, 0.63)
Regression equation:
Commitment = 1.68 + 0.62Identification
What analysis options are there for the following study?
Q - does organisational identification predict commitment to an organisation
H - stronger organisational identification will predict stronger commitment OVER AND ABOVE (ie controlling or adjusting for) other factors like job satisfaction, salary, age, and sex
M - questionnaire measures of all variables (salary = annual, before tax)
Multiple Regression
- allows the role of individual predictors to be assessed while adjusting for the role of other predictors and assessing the overall fit of the model
What does the output from this multiple regression analysis tell us
According to the first table, the overall model is significant
According to the first table, the model explains 47% of variance in the DV
According to the second table - salary and job satisfaction are significant positive predictors, as is organisational identification
Age and sex, however, do not significantly predict commitment
What does the coefficients table in a multiple regression analysis output tell us
Indicates whether an IV explains a unique part of the variance in the DV
What analysis options are there for the following study when you want to know what the contribution of additional predictors are to explaining organisational commitment
- what is the size and significance of their regression coefficients
- what do they contribute to r2
(Q - does organisational identification predict commitment to an organisation
H - stronger organisational identification will predict stronger commitment OVER AND ABOVE (ie controlling or adjusting for) other factors like job satisfaction, salary, age, and sex
M - questionnaire measures of all variables)
Hierarchical multiple regression
What does this output from a hierarchical multiple regression tell us
The model is significant at all 3 steps
Significance of adding sex, age, and job satisfaction - they do explain extra variance (5.7%)
So does adding salary at the final step (3.1%)
You predict that X will predict DV, over and above Y & Z
In a multiple regression which blocks will you enter the variables
In a hierarchical regression?
Multiple regression:
enter X, Y & Z in one block (will test unique effects of each)
Hierarchical regression:
enter X, then Y & Z IF want to see whether effect of X differs after adding Y & Z
otherwise enter Y & Z, then add X
Identification has been added last to the block in this hierarchical regression output - what does it tell us?
Indicates that identification explains an additional 7.9% of the variance in commitment
F(1, 114) = 17.55, p<.001
What does this hierarchical regression output tell us about OrganisationalID
It remains significant but the effect (coefficient) of identification gets smaller as further predictors are added
This is because the other factors can be correlated with one another (meaning they have overlapping variance) . Factor A may explain some of the variance in the DV by itself but not uniquely if that variance is also shared with B
If A and B overlap a lot then very little can be uniquely attributed to A or B, neither of them may be significant when entered together
How to include categorical predictors as well as continuous predictors in regression
- example: participant sex
- example: departments in an organisation (sales, development, IT)
Sex:
this is a categorical predictor with two levels
interpretation depends on the coding - if female = 1, and male = 2, then the regression coefficient will represent the effect on the DV of being male rather than female (if male=1, and female=2 then it would be the other way around)
Departments:
this is a categorical predictor with 3+ levels
the numbering of these is arbitrary and meaningless to analyse so we would use dummy variables (3 categories = 3 dummy variables)
this allows us to test the significance and explanatory role of the variable as a whole
as well as the effect of being in a specific category relative to a reference catefory (which is ‘dropped’ from the analyses)
What does the output of this regression analysis tell me about the categorical predictor of sex
Being male rather than female (if female=1 and male=2) increases commitment by 0.22 (rounding up) - it is a non-significant effect (p = 0.438)
What does the output of this regression analysis tell me about the categorical predictor of departments (which has 3 levels)
According to second table - adding department as a variable didn’t significantly improve model fit (r2)
The third table (coefficients) indicate the effect of being in that category relative to the reference category (IT in this case)
What is a cross-lagged analysis
Deals with the issue of causality by measuring BOTH predictor and outcome at two(+) timepoints
- you’re essentially looking at how organisational identification at T1 AND commitment at T1 effect commitment at T2
What is the fully cross-lagged model