Week 6 - Non-parametric & measures of association and hierarchy of evidence

Apart of Introduction to Evidence Based Practice and Research in Health Sciences at UniSA

What is the difference between correlation and correlation coefficient?

Correlation is a technique used to investigate the relationship between two numerical variables whereas correlation coefficient measures the relationship between two numerical measurements (which tells us the magnitude of relationship, direction of relationship, and bivariate distribution).

What is the difference between positive and negative correlation?

Positive is present when high values in one variable are associated with high values of another variable or vice versa and negative is present when one variable is associated with low values or other variables or vice versa.

How is correlation displayed and how can it be analysed (2 points)?

Scatterplots and can be analysed by looking if it’s linear or non-linear association and if the trend of association is positive or negative.

What are the two types of correlation coefficients?

• Pearson correlation coefficient ®

• Spearman correlation coefficient (r_s)

Pearson correlation coefficient ®

A correlation coefficient which measures the strength of linear association between two quantitative variables.

How is Pearson correlation coefficient interpreted?

On a scale from -1 to +1 with zero representing no correlation, more than 0 is more positive, and less than 0 is more negative.

Spearman correlation coefficient (r_s)

A correlation coefficient which measures the strength of association between two ranked variables.

Why is Spearman correlation coefficient used over Pearson’s in some data sets (3 points)?

• If the assumption is markedly violated

• If there is a monotonic relationship between variables

• If the sample is very small

Does correlation imply causation, true or false?

False (e.g. MMR vaccination does not imply/cause autism)

Regression analysis

A common way of estimating the relationship among the variables (e.g. Given the age of an individual, can we estimate/predict their income levels)

Linear regression

The most basic and common type of predictive analysis where at the centre of the regression analysis, is the task of fitting a single straight line through a scatter plot (known as a regression line).

What is the difference between parametric and non-parametric data?

Parametric assumes normal distribution, homogenous variance, and data sets are typically ratio or intervals whereas non-parametric is where there are no assumptions on distribution or variance relationships and data sets are typically ordinal or nominal.

What are the advantages of using parametric and non-parametric data?

Parametric data can draw more conclusions from the sample to the population whereas non-parametric data is more simple and less affected by outliers.

In non-parametric statistics, is population median or population mean is used?

Population median.

Sign test

A type of non-parametric test which tests the hypothesis concerning the median where if the null hypothesis is true, there is approx. an equal number of observations greater and less than the median.