(1) regression Flashcards
(20 cards)
What is the difference between explanatory variable and response variable?
- Explanatory variable is the independent variable and can be controlled.
- Response variable is dependent variable as it depends on explanatory variable.
Why may a linear regression model be suitable?
-If points lie in a straight line
What is the interpretation for the b value in the form y= 0.4x+2
The rate of increase in y for every unit change in x.
For every unit increase in x y will increase by 0.4
What is something that can make using linear regression models unreliable?
-If the value is larger than the data set this is a extrapolation.
What happens to the PMCC in coded data?
-PMCC does not change
What is a causal and spurious relationship?
- Causal is when the explanatory variable has a direct impact
- Spurious is when it is coincidental
What is an r value?
- also known as PMCC
- The strength of the linear relationship between two variables.
What would happen to the PMCC If a set of variables was added that does not go with the straight line?
-It would weaken the PMCC as it does not follow the relationship.
why is a model y=ax^n a good model for the relationship between two variables if r=1
-As r is close to one graph of log(y) and log(x), supporting relationship of t=ax^n
How do you carry out a zero correlation hypothesis test.
-Step 1 one tailed or two tailed. positive correlation?
(Divide sig level by two if it is a two tailed test
-Step 2, use sample size and sig level to find critical value.
-Step 3- H0: p=0
H1 p: less than or greater than p
- Step 4 see if r value falls in critical region. (greater than the critical value. Draw out number line.
- Step 5 if it falls in critical region reject H0
- Step 6 link it back to context.
What are the two ways to use logs for linear relationships?
type 1 -y=ax^n
y=ab^x
How to move from y=a+bx to coded version?
-sub in the coding formula for y and x
What is the particular rule for undoing codes for mean and standard deviation?
As mean is average code as normal
As standard deviation is measure of spread do not add or subtract
What is a justification of why you cannot use y=bx+c to work out value of x when y=10
-As the regression line is for y given x, to work out x use x on m regression line.
What is correlation?
-The nature of linear relationship b between two variables.
What do the different r values mean
r= -1 perfect negative relationship
r=1 perfect positive relation
r=0 perfect no linear relationship
How do you perform a hypothesis test for zero correlation for one tailed?
- State h0 and h1
- Use significance level and sample size to find critical value using PMCC table
- Draw a diagram and shade critical region between CV and H+
- If r value falls in critical region reject null hypothesis.
- Conclusion in context of the question.
How do you perform two tailed hypothesis test for zero correlation?
- Divide significance level by 2.
- DIagram will have h+ and h-
How would you articulate the effect on PMCC?
-Would weaken/strengthen the correlation as it goes against what PMCC shows
Why would a linear regression model be appropriate to describe f and d?
-because they lie on straight line.
