# Statistics

## 71 terms

how far off your expectation will be

average

most common

max - min

### resistant to outliers

if you take out the outliers, these do not change very much

### median

50% of data below, 50% data above

max-min/2

### emperical rule

68%= 1 standard deviation
95%= 2 standard deviation
99.7%= 3 standard deviation

far away

### measures of center

what you should expect

### variance

-standard deviation squared
- not resistant to outliers

### When is data unusual?

outside 2 standard deviations

### standard deviation

-always positive
-not resistant to outliers
-"average distance from the average"

### continuous

no breaks
(always something in-between)

### interval

has order, spacing and no true "0"
-doesn't start at zero
(body temp)

all

### census

how to measure population

### ratio

order, spacing, true "0"
starts from zero
(money in the bank)

### nominal

no logical order can be applied
(red blue yellow)

### qualitative

words, where you fit in
(can't be continuous)

not "all"

### quantitative

numbers, how many, how much
(both discrete and continuous)

### discrete

"has breaks"
nothing in the middle
(how many boys in classroom?)

### parameter

populations (better to have)

### ordinal

has order but no spacing
(G, PG, PG-13, R) ABCDF

### statistic

sample (easier to get)

### voluntary response

passionate or extreme call in
(not usually good)

relationship

### causality

one thing actually causes the other

### confounding

other stuff is involved

1. not truthful
2. don't know

### small sample

each one person is a large part of population

people are dumb

### order of question

two bad things, you choose the second
two good things, you choose the first

### non-response

when asked people choose to respond or not

### missing data

whole groups are left out

### self interest study

"domino's is america's favorite"
"Pizza hut is America's favorite"

All out lie

### Precise numbers

(exact) NOT ACCURATE

### observational study

do nothing
2) when impossible to do experiment
3) won't be unethical

### experiment

do something
apply, treatment
control other variables and eliminate confounding

-prospective
-retrospective
-cross-sectional

future

past

### cross-sectional

small slice of time

fake treatment

### blinding

subject does not know the treatment

### double blinding

subject and the data recorder do not know the treatment

### blocking

separate sample by "lurking variable"

### totally random experiment

no bias, chance of "bad" sample

### rigorous control experiment

lots of bias, "best" sample

### random sample

every individual has the same chance to be chosen

### simple random sample

each group has a chance of being chosen (if not random sample, can't be simple random)

### systematic

every nth individual

### convienance

easy, get a lot of overlapping stuff

### stratified

break into groups and choose some from each groups

### cluster

break into groups and choose all from some groups

### non-sample errors

we have poorly designed or executed an experiment

### variation

how much does it change

has a bell curve

### z-scores

lets us compare apples and oranges
Z=data-mean/STD

### percentile

# of data below/# of data x 100
P=B/T

25th percentile

50th percentile

75% percentile

Q3-Q1
(middle 50%)

### Outliers by IQR

High: 1.5 IQR rule above Q3
Low: 1.5 IQR rule below Q1

within -2 to 2

### unusual

above or below -2 or 2