L6 - Exploring Relationships

Question 1

Why do researchers explore relationships?

Answer 1

Researchers want know if there is a significant relationship between two variables

**OR** if there is a significant ****DIFFERENCE**** between one or more

Question 2

What are the requirement for drawing conclusions?

Answer 2

• You need to have equal sample size
• How reliable is the data gathered
  
  • Is the results given up to chance
  • Stats test needed to calculate significant value

Question 3

What values are important when performing statistical tests?

Answer 3

Confidence value (Significant / Alpha) - Observed relationship or difference between groups or treatment conditions (Taken from stats test)

Significants (p) - is the probability that the observed relationship or difference is due to chance or accidental

• Depends on size or degree of relationship
• Degree of variablility or dispersion owithin the sample
• Depends on the size of the sample

****Lower significance value → lower chance that the result is accidental****

**p = 0.05 is the threshold of significance**

Question 4

What are the realtionship tests?

Answer 4

• Correlation - are variable X and variable Y related
• Multiple regression - which combination of independent variables offers the most accurate prediction of a given dependent variable
• Chi Squared - shows weather there is an association between two categorial variables

Question 5

What are differences tests?

Answer 5

• T-tests measure the diffference between two groups (males and females) according to some continuous variable (Height)
• Pearsons R

Question 6

****What is Independent variable****

Answer 6

• Variable believed to affect the dependent variable
• What you change

Question 7

************What is dependent variable / outcome variable************

Answer 7

• The observation that is believed to be affected by the IV
• What you measure

Question 8

**What is the level of measurement of each variable**

Answer 8

• Nominal, ordinal or continuous data
• Level of dependent measure determines ************parametric or non-parametic test************

Question 9

What are parametric tests?

Answer 9

**Parametric tests can make certain assumptions about parameters of sampled populations**

• Fairly robust - tolerant of minor violations of assumptions
• Requires data to be
  
  • Normally distributed
  • Variants of comparison groups are equivalent
  • Measurements are on a true continous scale

Question 10

What are non-parametric tests?

Answer 10

**WHEN IN DOUBT USE NON-PARAMETRIC**

• Make no asssumptions about distrabutions
  
  • Chi Squared
  • Mann-whiney U
  • Kruskal-Wallis or ANOVA
• Analyse rank-order data rather than continuous data
• Less powerful (less likely to reject null hypothesis)

Question 11

What are the different types of correlation?

Answer 11

Question 12

What is the link between correlation and causation?

Answer 12

• Relationships can be causal
• But correlation doesnt always mean causation
  
  • Further theoretical basis is needed to assume causative effect
  • Is there a third party variables effecting the results

Question 13

How does pearson R work?

Answer 13

Question 14

How do you compute Pearsons R?

Answer 14

• Red is coefficient - -1 to +1 → this case its negative correlation
  
  • Positive = X Y vary in the same direction
  • Negative = X Y vary in different directions
  • Strength of the relationship
    
    • Small = 0.10→0.29
    • Medium = 0.30→0.49
    • Large = 0.5→1.0
• Blue is Significance (p) value - LOWER IS BETTER = less possibility for error
• N is number of valid cases

**Proportion of varience shared by X AND Y** (Example is ~34%)

= $r^2 * 100$