exam 3

Question 1

What does a significant “interaction effect” tell you? what type of ANOVA allows you to look for interaction effects between variables?

Answer 1

• That the two variables together give a stronger effect then the two variables alone. when together, the effects are stronger.
• Multifactorial ANOVA

Question 2

Explain the difference between correlation and causation. What type of study is needed to determine causation?

Answer 2

Correlation: they are related but one doesn’t cause the other
Causation: We can tell that one variable influences the other
- the study that has to be done is experimental approach

Question 3

what test should be used:

Does listening to music while studying (music, no music) influence how long you retain information (scalar)? Researchers wanted to control for any effect of differences in volume (dB)(scalar) in the analysis.

Answer 3

ANCOVA; volume is our covariate bc we don’t care about it itself

Question 4

What test should be used:
How do light levels ( 12 hrs per day or 16 hrs per day ) and nitrogen supplementation (nitrogen supplement vs. vehicle) together or independently influence fruit size (diameter in mm) in tomatoes? Independent samples. Parametric.

Answer 4

Multifactorial ANOVA

Question 5

repeated measures ANOVA

Answer 5

• only option for repeated measures

Question 6

multifactorial ANOVA

Answer 6

• there is more than independent variable
• usually description that they are looking for those interaction effects or individual effects

Question 7

MANOVA

Answer 7

• Multivariate ANOVA
• multiple dependent variables
• we worry they may be correlated to each other
• A->B
• B->A
  -C-> A&B

Question 8

ANCOVA

Answer 8

• there is a covariate
• there is a variable that they don’t care about by itself, just what it causes

Question 9

Pearson’s correlation

Answer 9

• model building technique
• parametric
• assumes linear shape
• sensitive to outliers

Question 10

Bivariate linear regression

Answer 10

• two variables and linear shape
• Contains one predictor

Question 11

multiple linear regression

Answer 11

• more than one predictor
• linear shape
• the more predictors, you will increase R-squared
• uses post-hoc test

Question 12

4 parameter logistic regression

Answer 12

• non-linear shape
• makes s-curve
• Has upper asymptote, Lower asymptote, Inflection point, and Slope at inflection point

Question 13

binary logistic regression

Answer 13

• variable trying to predict is not scalar
• can be categorical or scalar
• one or more predictors with one result
• creates a probit

Question 14

What kind of regression would be used:

A researcher wants to create a model to predict the growth rate of brine shrimp (scalar) based on temperature (scalar) and salinity (scalar)

Answer 14

Multiple linear regression (multiple predictors)

Question 15

Multifactorial means what?

Answer 15

There are more than one independent variable

Question 16

what is the parametric assumption of the Pearson’s correlation?

Answer 16

each variable is normally distributed

Question 17

What is the purpose of regression techniques?

Answer 17

To create a predictive model

Question 18

What is the calculated model for bivariate linear regression?

Answer 18

y=mx+b

Question 19

where do you find how much variation is explained in bivariate linear regression?

Answer 19

R-square value

Question 20

how do you know if bivariate linear regression explains a significant amount of variation? how do you report it?

Answer 20

• On ANOVA output, look to see if the p-value is less than alpha. if it is, then it is significant.
• ex. Yes. F=104.624, p<0.001

Question 21

what is a post-hoc test for?

Answer 21

to interpret which variable/s is significant

Question 22

when reporting R-square value, which one do we report?

Answer 22

the adjusted R-square

Question 23

problems with 4 parameter logistic regression?

Answer 23

• produces poor estimation
• mis-predicting

Question 24

what is a probit?

Answer 24

• probabilistic model; probability for each of the 2 outcomes
• will not always match up

Question 25

What is the purpose of regression techniques? How do they differ from correlations?

Answer 25

- Regression techniques model the relationship between a dependent variable and one or more independent variables (the predictors). - regression shows how much one variable changes another. - correlation shows how strongly things are related.

Question 25

You find that glyphosate (a pesticide component) exposure is significantly and positively correlated with the amount of vegetables consumed in the diet. How would you interpret this result? What would you need to do to evaluate possible alternative explanations?

Answer 25

Correlation does not determine cause and effect. - There are 3 options A causes B, B causes A or an outside source C Causes A and B Independently. - I would say the glyphosate and amount of vegetables consumed are not independent. - you would need an experiment as well as a post-hoc test to determine which predictor.

Question 26

given the following data, Write the equation for the bivariate linear regression reported: - Constant unstandardized B= 2.820 - mass unstandardized B= .024 - dependent variable=Wingspan

Answer 26

Wingspan= 0.024 (mass) +2.820

Question 27

given the following data, what is the calculated model and variation in wingspan predicted: - mass = 0.021 - shoulder width = 0.024 - constant= 2.255 - r-squared= 0.431 - adjusted r-squared= 0.410

Answer 27

calculated model: wingspan= 0.021(mass)+ 0.024(Shoulder width)+2.255 Variation in wingspan predicted: 0.410

Question 28

Why does explanatory power increase as a model becomes more complex? Why is this a problem?

Answer 28

- when you include more predictors it grows more specific. - Each predictor has error with it. With the r-squared increasing it initially is thought as a good thing, but the data is becoming too specific - more predictors= data is better at determining the noise

Question 29

Why is it important that researchers choose the appropriate shape of model they want to use before they look at the data?

Answer 29

- you should choose the shape first to avoid p-hacking. When looking at data, you are tempted to choose a model that gives the best results rather than reflecting the actual data. - it will also prevent overfitting.

Question 30

Autocorrelation

Answer 30

one dependent directly determines the other

Question 31

Omnibus test

Answer 31

looks at both together followed by a univariate test

Question 32

Spearman's correlation

Answer 32

- non-parametric - ranks points for each variable and calculating the difference between ranks