Title | Stat 100 Final Cheat Sheets - Google Docs |
---|---|
Author | Tyesee Michelle |
Course | Introduction to Statistics |
Institution | University of Maryland Global Campus |
Pages | 14 |
File Size | 2.2 MB |
File Type | |
Total Downloads | 57 |
Total Views | 159 |
Download Stat 100 Final Cheat Sheets - Google Docs PDF
Populationentirecollectionofobjectsor
➔ Meanarithmeticaverageofdata values ◆ **Highlysusceptibleto extremevalues(outliers). Goestowardsextremevalues ◆ Meancouldneverbelargeror smallerthanmax/minvaluebut couldbethemax/minvalue ➔ Medianinanorderedarray,the medianisthemiddlenumber ◆ **Notaffectedbyextreme values ➔ Quartilessplittherankeddatainto4 equalgroups ◆ BoxandWhiskerPlot
individualsaboutwhichinformationisdesired.
➔ easiertotakeasample ◆ Samplepartofthepopulation thatisselectedforanalysis
◆ Watchoutfor: ●
Limitedsamplesizethat mightnotbe representativeof population
◆ SimpleRandomSampling Everypossiblesampleofacertain sizehasthesamechanceofbeing selected
ObservationalStudytherecanalwaysbe lurkingvariablesaffectingresults ➔ i.e,strongpositiveassociationbetween shoesizeandintelligenceforboys ➔ **shouldnevershowcausation
➔ Variancetheaveragedistance squared n
s2x = ◆
n 1
sx2 getsridofthenegative
values ◆ unitsaresquared ➔ StandardDeviationshowsvariation aboutthemean
s =
√
n
∑ (x i x) i=1
n 1
2
◆ highlyaffectedbyoutliers ◆ hassameunitsasoriginal data ◆ finance=horriblemeasureof risk(trampolineexample)
controlled;cangivegoodevidenceforcausation
i=1
ExperimentalStudylurkingvariablescanbe DescriptiveStatisticsPartI ➔ SummaryMeasures
∑ (xi x)2
DescriptiveStatisticsPartII LinearTransformations
➔ Range= X maximum X minimum ◆ Disadvantages:Ignoresthe wayinwhichdataare distributed;sensitivetooutliers ➔ InterquartileRange(IQR)=3rd quartile1stquartile ◆ Notusedthatmuch ◆ Notaffectedbyoutliers
➔ Lineartransformationschangethe centerandspreadofdata ➔ V ar(a + bX) = b2 V ar(X) ➔ Average(a+bX)=a+b[Average(X)]
➔ EffectsofLinearTransformations: ◆ meannew = a+b*mean ◆ mediannew = a+b*median ◆ stdev new = | b| *stdev ◆ IQRnew = | b| *IQR ➔ Zscorenewdatasetwillhavemean 0andvariance1 z = XS X EmpiricalRule ➔ Onlyformoundshapeddata Approx.95%ofdataisintheinterval: (x 2 sx , x + 2 sx ) = x + / 2 sx ➔ onlyuseifyoujusthavemeanandstd. dev. Chebyshev'sRule ➔ Useforanysetofdataandforany numberk,greaterthan1(1.2,1.3,etc.) 1 ➔ 1 2
Skewness ➔ measuresthedegreeofasymmetry exhibitedbydata ◆ negativevalues=skewedleft ◆ positivevalues=skewedright ◆ if |skewness | < 0.8 =don'tneed totransformdata MeasurementsofAssociation ➔ Covariance ◆ Covariance>0=largerx, largery ◆ Covariance5and
2
ifpopulationisnormallydistributed, ncanbeanyvalue anypopulation,nneedstobe ≥ 30
populationproportionisintheinterval…
➔ mean( x) = μ ➔ variance (x) = σ2 /n ◆
=
numberof successesinsample samplesize
➔ ➔ Wearethus95%confidentthatthetrue
mean)
➔
x n
StandardErrorandMarginofError
B. OneSampleMean Forsamplesn>30 ConfidenceInterval:
*Stataalwaysusesthetdistributionwhen computingconfidenceintervals HypothesisTesting ➔ NullHypothesis: ➔ H 0 ,astatementofnochangeandis assumedtrueuntilevidenceindicates otherwise. ➔ AlternativeHypothesis: H a isa statementthatwearetryingtofind evidencetosupport. ➔ TypeIerror:rejectthenullhypothesis whenthenullhypothesisistrue. (consideredtheworsterror) ➔ TypeIIerror:donotrejectthenull hypothesiswhenthealternative hypothesisistrue.
➔ Ifn>30,wecansubstitutesfor σ sothatweget:
ExampleofSampleProportionProblem
ExampleofTypeIandTypeIIerrors DeterminingSampleSize ︿
n =
︿
(1.96) 2p(1 p) e2
︿
➔ Ifgivenaconfidenceinterval, p is themiddlenumberoftheinterval ➔ Noconfidenceinterval;useworst casescenario
Forsamplesn...