Fichas
RM Lecture 1
Fichas del conjunto (35)
- 1What is regression?A way of predicting data points you haven't measured. Can be used to create a model of the relationship between two variables.
- 2What is the difference between correlation and regression?Correlation has no predictions and regression is more powerful.
- 3What is simple regression?Regression that predicts an outcome variable from a predictor variable.
- 4What is multiple regression?Regression that predicts an outcome variable from several predictor variables.
- 5What ar ethe outcome and predictor variable also known as?Outcome - dependent
Predictor - independent. - 6What is the method of least squares used for?To find the line that fits the data best.
- 7How does SPSS calculate the method of least squares?It minimises the squared difference between the data and the line.
- 8What is the slope of the line also called?Gradient.
- 9What is the point where the slope crossed the y-axis called?Intercept.
- 10What is the equation of a straight line?y=mx+c
(m = gradient and c = intercept) - 11What are the gradient and intercept denoted as in multiple regression?Intercept = b0
gradient = b1, b2... - 12What is the line of best fit?The line with the least amount of difference between the data points and the line.
- 13What do we compare all line fits to?The null hypothesis (a horizontal line) meaning that y has no dependence on x.
- 14What is the difference between the line and the data point called?Residuals.
- 15Why do we have to square the errors?So that the positives and negatives don't cancel out.
- 16Why can we not tell if the line of best fit is representing a true relationship?Because SPSS can draw a line of best fit through anything.
- 17How can we tell if a line of best fit is actually showing a true relationship?By comparing it to the null hypothesis.
- 18What is the total sum of squares (SST)?Total sum of squares of all the data points in relation to the null hypothesis line.
- 19What is the sum of squares of the residuals (SSR)?Total sum of squares of all the data points in relation to the line of best fit.
- 20If there's a relationship, should SST or SSR be smaller?SSR.
- 21What is the model sum of squares (SSM)?SST-SSR
- 22What does SSM tell us?How much better than the null hypothesis the model is. Bigger SSM = better fit.
- 23Whats wrong with SSM?It changes depending on the units.
- 24How do we get SSM to tell us what fraction of the variance is accounted for by the model?SSM/SST (bigger % = better fit)
- 25What is SSM/SST called?R2 - the coefficient of the determination.
- 26What is the equation for R2?(SST-SSR)
SST - 27What are the 3 main points about R2?1. Transferable (independent of units)
2. Intuitive (between 0-1)
3. Not the same as a p-value. - 28What is R2?The standardised goodness of fit.
- 29What do you need to get a p-value?F-statistic.
- 30What si the F statistic?Its the average amount of variance explained by the fit.
- 31Sums of squares are total values that can be expressed as averages… what are these called?Mean squares (MS)
- 32What is the equation for F?MSM/MSR
(MSM = mean square of the model)
(MSR = mean square of the residual) - 33What is B0 also known as?Constant.
- 34What conditions must be met to run a regression?Outcome = continuous
Predictors = continuous/discrete
Non-zero variance - predictors shouldn't have zero variance
Independence of data points
Linearity
Homoscedasticity - for each value of the predictors the variance of the errors should be constant (cloud)
Independent errors - For any pair of observations , error terms should be uncorrelated (independent measures) - Durbin-Watson
Normally-distributed errors (Zresid - residuals agains Zpred - predicted variable) - should be no triangles, can also look at histograms. - 35In regression do we do normality tests before or after?After.
Información
- Descripción
- Fichas sobre RM Lecture 1, creado por hannahbevans el 20/12/2014.
- Creado por
- hannahbevans
- Visitas
- 3
- Valoración