STAT C8 Study Guide - Final Guide: Test Statistic, Percentile, Table Table
3112 Confidence Intervals 13 ,
13.1 ,B. 2
.one sample gives One estimate
.estimate =parameter terror
.what is this error ?cant get another sample a
.large random sample :sample from that group WITH
bies
BOOTSTRAP
def bootstrap (original ,label ,replicate :
medians =make .away c)
for iin np .arange (replicate ):
bootstrap -samp =Original .sample ()
re sampled ME D= percentile (so ,boot .column (e)
medians =hp .append (medians
,resampkd MED
return medians
•95% confidence interval
.actual should be in between left =Percentile 12.5 ,medians
fight
percentile (97.5 ,medians
.now calculate 958 confidence intend 1100 times &see no
many times its accurate
3114 Interpreting Confidence level
3/16
Variation =(state
Std :how far from Mean mfefnof square
→bulk of data within few stds
3/19 The Normal Curve
•average ±zsDs is at least I-Is }irregardless of shape of the population
AT Least
#
Norma Distribution
2SD :75% bP§,
3:99.7
3SD :88.88%
€39934508
Chlbyshev 's Bounds
standard units
2- Imf
SD
2:how many
PD
2=0 =average
histogram :If belt shaped
center →center
points of inflation :goes down
CLT : it Mrgel random sample ,
then regardless Of distribution
probability distribution of the sample Sumino is normal
3123 -
Population Sd
5¥
=SD of
sample means
distribution of pop :not normal
distribution Of sample : normal
1.) Large sample
2.) draw samples Of that size
3.) copute averages ⇒distribution of average
SD =PISA
Famfle size
95% :2sD5 Population and sample are 2SD 's away from each other
Document Summary
13. 1: 2 one sample estimate gives parameter. One estimate terror what is this error cant get another sample a large random sample : sample from that group. Bootstrap def bootstrap medians ( original , label replicate make for i in np . a bootstrap re sampled medians return medians away c) replicate ) range. Me d= percentile sample ( so boot column (e) = hp . append ( medians resampkd med: 95% confidence interval actual should be in between left. = percentile fight percentile ( 97. 5 , medians. Is } irregardless of shape of the population. Clt : it mrgel random sample , then regardless of distribution probability distribution of the sample sumino is normal. 5 sample means distribution of pop : not normal distribution. 2. ) categorical: association trends t 1- associations measure in. : any shape ( doesn"t have to be linear like trends. Point of averages a relation coefficient r measures linear.
