PS296 Lecture Notes - Lecture 10: Standard Deviation, Sampling Distribution, Null Hypothesis
Lec 10 – Correlation Continued
Characteristics of the sample that affect r:
1) range restrictions:
-restrictions on the range over which x and y can very usually reduces r
2) non-linear relationships
-sometime a curved line better suits the data
3) heterogenous subsamples
-the sample could be divided into two distinct sets of data on the basis of some other variable
-correlation coefficients, like other stats, are affected by sampling error
-correlation coefficients calculated with sample data can deviate from true correlations in the
population
BUT, how much deviation is normal and/or expected?
rho (ρ)
-we use r to estimate rho
-the correlation coefficient for the population
-we will know its a correlation (as opposed to z test) if question specifies two variables and asks to
figure out a relationship
General steps involved in hypothesis testing:
Document Summary
Characteristics of the sample that affect r: range restrictions: Restrictions on the range over which x and y can very usually reduces r: non-linear relationships. Sometime a curved line better suits the data: heterogenous subsamples. The sample could be divided into two distinct sets of data on the basis of some other variable. Correlation coefficients, like other stats, are affected by sampling error. Correlation coefficients calculated with sample data can deviate from true correlations in the population. But, how much deviation is normal and/or expected? rho ( ) We will know its a correlation (as opposed to z test) if question specifies two variables and asks to figure out a relationship. Step 2: calculate covariance of x and y. Step 2: calculate standard deviation x = 6. 178 y = 1. 751. Step 3: to check whether this is significant, we need to find critical value for alpha = . 05 for our degrees of freedom.