AP Stat Test 2

What is blocking: grouping already selected subjects into subgroups(blocks) and then randomly assigning each block. Only do if you think groups will react different to T’s. Done after random selection.

Purpose of blocking: To reduce variability so that differences can be more easily seen

Direct control: To reduce variability so that differences can be more easily seen

Completely randomized experiment design: most common type of experiment. Randomly assign the subjects to treatments (hat!)

Matched pairs experiment: either give both treatments to a single individual. Randomize the order. Other option is to match subjects based off similar characteristics and randomize treatments.

What is the point of matched pairs experiments: To reduce within subject variability

Control group: When one group receives no treatment. Only do if experiment is comparing one treatment to nothing.

Placebo: When one of the experimental groups you randomly assign to appears to be identical to another treatment group (in appearance, taste, feel, etc.), but it actually contains no active ingredients. It’s like a secret control group

Single blind: When the subjects don’t know which treatment they are receiving, but the people measuring the results do (like a nurse, say).

Double blind: When neither the subjects nor the person measuring the results knows which treatment a subject is receiving. (but the experimenter knows!!)

What chart do we have for numerical data?: Dot plot

What chart do we have for categorical data?: Bar chart

In a mosaic chart what does each bar’s WIDTH represent?: Its frequency.

How to use a list of random numbers to do stuff like random selection or random assignment (32310     66102     98546     77531     12366     43488     93990):  Choose two numbers at a time. 01-30 represent people. Ignore 31-99,00. Ignore repeats.  Continue choosing until you have three people.

How to decide # of classes for a histogram: # of classes = sq root of # of observations

How to figure out class width/intervals: class width = high - low / # of classes

Another name for left skewed: negatively skewed

Another name for right skewed: positively skewed

How do you fix unequal intervals?: divide the frequency by the width, this is called density.

Things that make graphs go bad:  SCALING badly is the BIGGEST way that good graphs can go bad. For example, not starting at zero can be naughty. It exaggerates differences. (Not true with scatter plots)

What are scatter plots used for?: Bivariate data (numerical data)