Quiz 9

Question 1

When does it make sense to create a histogram for your R data?

Answer 1

• you have only one variable AND
• the variable has “continuous/numeric” data AND
• you want to check if data are normally distributed

Question 2

What would you type in R to make a histogram of a particular variable (ex. column “Mean_RT”) (mean reaction time)

Answer 2

ggplot(ldt_df, aes(x=Mean_RT))+
geom_histogram(bins=13, color= ‘black’, fill= ‘light blue’)

Question 3

What is the function to save a ggplot?

Answer 3

ggsave(‘my_histogram.png’,h=5, w=7, units=‘in’)

Question 4

What does the ‘aes’ function do?

Answer 4

Allows you to assign what will go on X and Y axes
ex. aes(x=Mean_RT)

Question 5

How many variables do histograms need and how do you assign the variable/s to an axis?

Answer 5

• only one (numeric) variable
  ex. aes(x=“Mean_RT”)
• this assigns the “Mean reaction time” variable to the x axis

Question 6

What do you need to check in your data when you are assigning a variable to an axis when making a histogram?

Answer 6

That it is numeric

Question 7

Besides assigning the variable using the ‘aes’ function in R, what is needed to make a histogram?

Answer 7

Need to tell R we want a histogram plot

Question 8

What are the different plot options we are using in ggplot?

Answer 8

• histogram
• box plot
• scatter plot
• bar plot

Question 9

How do you specify what kind of plot you want in R?

Answer 9

use the ‘geom’ function
ex. geom_histogram() or geom_boxplot()

Question 10

What are “bins” in R (ggplot)?

Answer 10

bars in histogram

Question 11

How can you tell R to make the bins of your histogram different colors and fills?

Answer 11

ex. geom_histogram(bins=13, color=‘black’, fill=‘lightblue’)
(this also sets number of bins to be shown)

Question 12

When is it better to use mean?

Answer 12

When the distribution is normal

Question 13

When would you choose to use a box plot?

Answer 13

You have 2 variables and one variable numeric (“continuous”) and one categorical
- categorical variable has more than one group/distribution
- interested in seeing whether groups/distributions are different from each other

Question 14

In boxplots, which variable goes on which axis?

Answer 14

x-variable is categorical
y-variable is continuous

Question 15

How do we approach the data when doing descriptive statistics?

Answer 15

with the intention of summarizing the characteristics of the data
ex. calculate mean of a continuous variable (like avg duration of English unstressed vowels)
count how many times something occurs in the dataset (ex. frequency of a word in a corpus/book)

Question 16

How do we approach the data when doing inferential statistics?

Answer 16

with some predictions in mind, running tests to verify the predictions
ex. noticed that the mean duration of stressed vowels was different from that of the unstressed vowel in our sample
*so, we make a prediction (“infer”) that the 2 means are significantly different from each other

Question 17

What is done during inferential statistics to verify whether predictions are valid?

Answer 17

run a significance test

Question 18

What is an example of a null hypothesis?

Answer 18

ex. there is NO significant difference between stressed and unstressed vowels in terms of duration
(assumes no relationship between 2 variables and that controlling one variable has no effect on the other)

Question 19

What is done in a study that has a null hypothesis?

Answer 19

Make every effort to prove Null hypothesis wrong

Question 20

What is the abbreviation for a Null hypothesis?

Answer 20

H0

Question 21

What is an example of an Alternative Hypothesis and how is it abbreviated?

Answer 21

“There is a significant difference between stressed and unstressed vowels in terms of duration”
Alternative Hypothesis (H1)

Question 22

Why is a null hypothesis a good way to begin a study?

Answer 22

Because it is not starting with a bias (ex. starting with something you want to prove and setting about trying to prove it)

Question 23

What is the question we want to ask if we have data that our null hypothesis might not be true?

Answer 23

How thin do we want that probably to be so we disbelieve our H0?
ie. what is the probability of us having the data we have and also having our null hypothesis be true?
ex. if hypothesis is there is no difference between mean salaries of carpenters and electricians, if there is data that shows a large difference in their salaries, the probability would be very low.

Question 24

What is the standard threshold (=alpha level) in social science and related fields when it comes to what probability we want to reject a null hypothesis?

Answer 24

0.05 (aka “p-value”, 5%)
always between 0 and 1

Question 25

What happens if you get a p-value less than 0.05? What is this called?

Answer 25

• decide that it is very unlikely to see such a difference (if there was no actual difference between two means)
• thus, only possible alternative is that our beginning assumption was wrong; therefore, the 2 means must be different from each other
• “accepting the H1 by means of rejecting the H0”

Question 26

What is it called when you don’t disprove the H0?

Answer 26

“fail to reject the H0”, can’t accept the H1

Question 27

If a study says something is “significant”, what measurement are they using?

Answer 27

less than 0.05 p value

Question 28

What does p-value stand for?

Answer 28

probability value
must be BELOW 0.05…can’t be 0.05 to be statistically significant (to disprove null hypothesis)

Question 29

Can you PROVE a null hypothesis?

Answer 29

No, you can only reject a null hypothesis or fail to reject it
ex. p-value is 0.1 (0.05 or more), fails to reject null hypothesis

Question 30

What are the steps in performing a hypothesis test?

Answer 30

(see slide 17 for flow chart)
1. look at the data type (continuous/categorical); compute summary stats (mean, median, counts, etc)
2. choose the test based on:
- how many DV’s and IV’s you have
- types of your DV’s and IV’s
3. check whether your data meets the assumption for a parametric test
- if it does, run the parametric test
- if it doesn’t, run the non-parametric test
4. See if you can reject the H0
- if yes, can accept your H1 to conclude that you see a significant effect/difference, etc
- if not, can conclude there was no significant effect/difference

Question 31

What are parametric statistics (vs. non-parametric)

Answer 31

• parametric (when data normally distributed) ex. mean—more powerful, reliable than non-parametric, results more robust and can be reproduced
• non-parametric (can work when data not normally distributed) ex. median—can still be useful if you can’t use parametric because your data isn’t normally distributed

Question 32

Memorize flow chart slide 17 (significance testing slides)

Answer 32

Need to know for exam

Question 33

What is the purpose of descriptive statistics?

Answer 33

to make a summary of the data we have at hand
to explore existing data for patterns

Question 34

What are the types of data involved in descriptive statistics?

Answer 34

1. ratio— A x B, A/B
2. interval— A + B, A-B
3. ordinal— A>B or A<B
4. nominal—A doesn’t equal B

Question 35

What is nominal data? Examples?

Answer 35

When values are simply names or labels—least precise and informative level of measurement
ex. speaker of a language can be:
- native or non-native
- male or female
ex. languages can be
- English vs. French

Question 36

When making a boxplot, which kinds of variables go on which axis?

Answer 36

x-axis is categorical variable
y-axis is continuous/numeric

Question 37

What would you type in R to create a subset of your data, for example only keeping the “short” and “long” values from the Length_type column?

Answer 37

subset_df<-filter(ldt_df,Length_type==‘short’ | Length_type==‘long’)

Question 38

What does this character | mean in R and what is it called?

Answer 38

A pipe
allows us to ask R to keep the rows in which the both the variable values that are listed are present

Question 39

When would you do a scatterplot?

Answer 39

• you have 2 variables
• both variables have “continuous” data
• you are interested in checking if there’s a relationship between 2 variables (linear/non-linear, positive/negative, etc.)

Question 40

When making a scatterplot, which variable goes on which axis?

Answer 40

x axis- the “predictor” variable
y axis- the response/dependent variable

Question 41

What do you type in R to create a scatterplot?

Answer 41

geom_point()

Question 42

What is a paired t-test and when is it used?

Answer 42

• compares means of 2 related groups to check for significant difference between them
• considered paired because each observation in one group is directly related to an observation in the other group (ie. both groups tested under 2 conditions/pre-test-post-test)
• assumes normal distribution

Question 43

What is an unpaired t-test?

Answer 43

• compares means of 2 independent groups to determine if there is a significant difference between them
• groups not related or matched
• assumes normal distribution

Question 44

What is a paired U-test?

Answer 44

• alternative to paired t-test when distribution not normal
• non-parametric (uses median values)
• used for pre-test-post-test experiments
  H0 ex. Median difference between paired samples is zero (no difference)

Question 45

What is an unpaired U-test and when would you use it?

Answer 45

• non-parametric, used when data not normally distributed or small sample size
• compares 2 independent groups to evaluate difference between distribution of a dependent variable (ex. scores, measurements)
• variable of interest is either ordinal or interval/ratio that doesn’t meet parametric assumptions
  ex. Group A—teaching method 1, Group B—teaching method 2, look at difference between scores after teaching

Question 46

What is the ANOVA and when is it used?

Answer 46

• Analysis of Variance
• compare means of 3+ groups, otherwise like t-test
• looks at variability inside groups and differences between groups
• ex HO: all 3 groups means are equal
• ex. 3 fertilizers for plants—plants growth measured, tests to see if significant difference between them