MR Chapter 4

Question 1

What happens when you add a 4th variable?

Answer 1

The regression coefficients change in magnitude as we add new variables to the multiple regression equation because of - common causes or

• intervening variables.

Question 2

Why will common causes change the magnitude of a regression coefficient?

Answer 2

if a common cause is included, the regression coefficients will change from those found when such a variable is excluded from the regression, but R² will be the same.

Question 3

Why does an intervening variable change the magnitude of the coefficient?

Answer 3

because the regression coefficients focus only on direct effects.

this change in magnitude does not constitute a serious error in the analysis, unlike adding common causes which do seriously change things!

Question 4

should we fixate on R²? what is suspicious?

Answer 4

No, don’t focus on it and don’t be tempted to magnify it by adding variables, just add relevent vriables into the multiple regression.

Be suspicious if you see R² above .50.

Question 5

is prediction good?

Answer 5

Not as good as explanation.

prediction is scientifically less appealing, doesn’t allow you to think in terms of theory, interventions, policy or status quo.

don’t pretend to predict and then try to explain.

Question 6

what do you do when you want to predict and not explain?

Answer 6

If you want to PREDICT not explain, you want to maximize R2

Question 7

what do you do when you want to explain?

Answer 7

• Include any likely common causes
• Refreain from using irrelevant variables
• Use theory
• The higher the r² the better but it’s not the most important thing necessarily if we are doing MR for explanation.

Question 8

when do you include common causes?

Answer 8

To interpret regression coefficients as the effects of one variable on another

Question 9

why might adding a fourth IV decrease the effects of one variable on another?

Answer 9

If it’s a common cause we get smaller, more accurate estimates of effects.

Question 10

when you see a data output what are 4 things to consider?

Answer 10

• The metric of the variables (Whether to use b or ß)
• Their distribution/descriptives (Whether comparable across years?)
• Relationships between the variables. (What might affect what, what might be significant?)
• (If NELS data) N for each variable (implications for dfs) (These will be reduced when more data is included and accounted for.)