Bivariate Analysis- Austin

What is bivariate analysis?

analyzing the relationship between 2 different groups

What is the proper use of p-values?

for detecting “differences”

• reject null hypothesis: statistical significance

• fail to reject null hypothesis: no statistical significance

What are the misuses of p-values?

• p-value is not the probability the null hypothesis is true

  • #1 MISUSE= for PROBABILITY!!! DON’T DO THIS

• p-value should not be used alone

  • should have other supporting evidence

Does statistical significance mean clinical significance?

no

What are independent and dependent variables?

independent- values do not depend on any other variables in the scope of the experiment

dependent- values depend on or are affected by other variables by some rule or relationship

What are continuous variables?

can assume an INFINITE # of real values within a given interval

• any variable that can be measured

What are discrete variables?

can assume a FINITE # of real values within a given interval

• basically things that can be categorized/counted

What type of data belongs to continuous variables?

• interval

• ratio

What type of data belongs to discrete variables?

• nominal

  • dichotomous

• ordinal

What type of data is dichotomous?

nominal data

What is an example of dichotomous variables?

• yes/ no

• pass/ fail

What are the 3 steps to choosing the right statistical test?

1. how many groups of independent variables are there? 2? 3?

2. If there are 2 groups—- are they independent or dependent?

3. What is the nature of the outcome variable? AKA what type of data is it? Continuous? Nominal?

Contrast parametric and nonparametric data.

• Parametric	Non-parametric
  Assess mean	Assess median
  More powerful	Less powerful
  Normal	Skewed/ no shape
  Results significantly effected by outliers	No seriously effected by outliers
  Interval/ratio	All types of data

What are the 4 assumptions for parametric tests?

1. normality

2. independence

3. no outliers

4. equal variance

A t-test requires _____ independent samples or groups.

2

An ANOVA test requires _____ independent samples or groups.

3+

Both a t-test and ANOVA test have a ______________ outcome.

(continuous or discrete)

continuous

Is a t-test and ANOVA classified as parametric or non-parametric data?

parametric

What is the relationship between a t-test and ANOVA test? FYI

Basically the same thing HOWEVER a t-test has 2 independent groups and an ANOVA has 3 or more independent groups.

Chi-square tests are usually used to generate what?

a hypothesis

Chi-square tests require ____ or more ____________ samples or groups.

Chi-square tests require 2 or more independent samples or groups.

A chi-square outcome variable is ______________.

(continuous/discrete)

discrete (nominal or dichotomous)

A Fisher’s Exact test is EXACTLY like Chi-square, but used if we have <____ counts for any outcome.

5

A paired t-test differs from a regular t-test in what way?

paired t-test: two dependent samples or groups

• Ex: using the same subjects to test weight before and after exercise

t-test: two independent samples or groups

• using two separate groups of subjects, one no exercise and the other exercise

Just like a regular t-test, a paired t-test has a ______________ outcome variable and is _____________ data.

(continuous/discrete and parametric/nonparametric)

Just like a regular t-test, a paired t-test has a continuous outcome variable and is parametric data.

How many groups are required for a McNemar test?

2 dependent sample groups

True or False: A McNemar test has a continuous outcome variable.

false- dichotomous

Is McNemar parametric or nonparametric?

nonparametric

What are the requirements for a Mann-Whitney U test?

• 2 independent samples or groups

• DISCRETE… SPECIFICALLY ORDINAL outcome variables

• nonparametric

For Mann-Whitney U tests would you assess the mean or median?

median (nonparametric)

A Kruskal-Wallis Test is EXACTLY like a Mann-Whitney U test except for what?

Kruskal- Wallis test is for 3+ groups of independent variables, while Mann-Whitney U test is only for 2 groups

Does correlation mean causation?

NO!

What test is usually used to test whether the means of 2 populations are different?

t-test

What test is usually used to test whether two categorical variables are correlated?

chi-square

What 2 tests are usually used for pre/post testing?

• paired t-test

• McNemar

When is post-hoc testing necessary? What tests are they used in

when there are 3+ independent/ dependent variables

• used in ANOVA and Kruskal-Wallis

PRACTICE:

If I had the following dichotomous outcome, would I use a chi-square test or Fisher’s test for statistical analysis?

	Exercise Group	Comparison Group	Total
At Goal	7	2	9
Not at Goal	3	8	11
TOTAL	10	10	20

Fisher’s Test because there are outcomes with a count of <5

PRACTICE:

Which of the following tests compares median values?

a. t-test

b. paired t-test

c. Kruskal- Wallis Test

d. ANOVA

c. Kruskal-Wallis test is the only one with non-parametric data and we would use median