Download presentation

Presentation is loading. Please wait.

Published byAlexandra Christie Modified over 2 years ago

1
SADC Course in Statistics Correlation & the Coefficient of Determination (Session 04)

2
To put your footer here go to View > Header and Footer 2 Learning Objectives At the end of this session, you will be able to understand the meaning and limitations of Pearsons coefficient of correlation (r) describe what is meant by the coefficient of determination (R 2 ) and how it relates to r for a simple linear regression model derive the value of R 2 using results of an analysis of variance.

3
To put your footer here go to View > Header and Footer 3 What is Correlation? The term correlation refers to a measure of the strength of association between two variables. If the two variables increase or decrease together, they have a positive correlation. If, increases in one variable are associated with decreases in the other, they have a negative correlation.

4
To put your footer here go to View > Header and Footer 4 Linear Correlation (r) For two quantitative variables X and Y, for which n pairs of measurements (x i, y i ) are available, Pearsons correlation coefficient (r) gives a measure of the linear association between X and Y. The formula is given below for reference.

5
To put your footer here go to View > Header and Footer 5 Linear Correlation (r) If X and Y are perfectly positively correlated, r = 1 If there is absolutely no association, r = 0 If X and Y are perfectly positively correlated, r = -1 Thus -1 < r < +1 The closer r is to +1 or -1, the greater is the strength of the association.

6
To put your footer here go to View > Header and Footer 6 Possible values for r

7
To put your footer here go to View > Header and Footer 7 Coefficient of Determination It is often difficult to interpret r without some familiarity with the expected values of r. A more appropriate measure to use when interest lies in the dependence of Y on X, is the Coefficient of Determination, R 2. It measures the proportion of variation in Y that is explained by X, and is often expressed as a percentage.

8
To put your footer here go to View > Header and Footer 8 Using anova to find R 2 Anova (for 93 rural female headed HHs) of log consumption expenditure versus number of persons per sleeping room is: Sourced.f.S.S.M.S.FProb. Regression14.890 21.90.000 Residual9120.3420.2235 Total9225.2310.2743 R 2 = Regre. S.S. / Total S.S. = 4.89/25.23 = 0.194

9
To put your footer here go to View > Header and Footer 9 Interpretation of R 2 From above, we can say that 19.4% of the variability in the income poverty proxy measure is accounted for by the number of persons per sleeping room. Clearly there are many other factors that influence the poverty proxy since over 80% of the variability is left unexplained!

10
To put your footer here go to View > Header and Footer 10 Relationship of R 2 to r When there is just one explanatory variable being considered (as in above example), the squared value of r equals R 2. In the above example, value of r = - 0.194 = - 0.44 The negative value is used when taking the square root because the graph indicates a negative relationship (see next slide).

11
To put your footer here go to View > Header and Footer 11 Plot of poverty proxy measure vs. persons per sleeping room

12
To put your footer here go to View > Header and Footer 12 Benefits of R 2 and r r is useful as an initial exploratory tool when several variables are being considered. The sign of r gives the direction of the association. R 2 is useful in regression studies to check how much of the variability in the key response can be explained. R 2 is most valuable when there is more than one explanatory variable. High values of R 2 are particularly useful when using the model for predictions! (More on this later!)

13
To put your footer here go to View > Header and Footer 13 Limitations of r Observe that seemingly high values of r, e.g. r=0.70, explain only about 50% of the variability in the response variable y. So take care when interpreting correlation coefficients. a low value for r does not necessarily imply absence of a relationship – could be a curved relationship! So plotting the data is crucial! Tests exist for testing there is no association. But depending on the sample size, even low values of r, e.g. r=0.20 can give significant results – not a very useful finding!

14
To put your footer here go to View > Header and Footer 14 Limitations of R 2 Note that R 2 is only a descriptive measure to give a quick assessment of the model. Other methods exist for assessing the goodness of fit of the model. Adding explanatory variables to the model always increases R 2. Hence in practice, it is more usual to look at the adjusted R 2 The adjusted R 2 is calculated as 1 – (Residual M.S./Total M.S.) As with R 2, the adjusted R 2 is often expressed as a percentage.

15
To put your footer here go to View > Header and Footer 15 Practical work follows to ensure learning objectives are achieved…

Similar presentations

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google