Page 7 - Regression Guideline

Gecng"your"data"ready"for" analysis"
Step&Two:&Standard&DeviaDon&
&
The"standard"deviaJon"in"sales"price""(i.e.,"the"amount"the"price"of"a"property"varies" around"an"average"sales"price"of"the"selected"properJes)"is"then"calculated"for"each"of"the" selected"properJes."Prior"to"esJmaJng"the"regression"model,"these"standard"deviaJons" (see"the"next"secJon"explaining"the"use"of"standard"deviaJons"to"idenJfy"“outliers”"in" more"detail)"are"used"to"idenJfy"and"remove"properJes"with"extremely"high"or"low"sales" prices"relaJve"to"the"average"price"to"avoid"including"properJes"that"unnecessarily"inflate" the"variaJon"being"modeled"and"which"are"likely"too"dissimilar"to"the"appraised"property" to"be"relevant."
Sidebar:&What&is&a&Standard&DeviaDon?&
&
The"standard"deviaJon"is"a"staJsJcal"measure"of"the"amount"of"variaJon"in"a"sample."As" such,"it"is"o^en"referred"to"as"a"measure"of"“dispersion”."It"reflects"how"widely,"on" average,"scores"vary"around"a"mean."Standard"deviaJon"scores"provide"informaJon"on" cases"in"a"sample"to"help"idenJfy"those"that"are"very"far"above"or"below"the"average"score" or"mean."Outliers"are"important"to"idenJfy"because"they"can"influence"the"mean"so"that"it" is"no"longer"an"accurate"measure"of"the"sample"average.""
"
For"example,"let’s"say"that"for"a"group"of"100"properJes"the"mean"sales"price"is"$600,000." Despite"this"mean"price,"the"sales"prices"of"the"majority"of"the"properJes"clusters"around" $500,000."Within"the"sample"there"are"5"properJes"with"unusually"high"relaJve"sales" prices"of"$1,000,000,"or"more."There"could"be"many"reasons"why"these"properJes"were" valued"so"much"higher"than"the"majority"of"the"remaining"properJes."Regardless,"inclusion" of"their"sales"prices"with"the"rest"of"the"properJes"inflates"the"mean"sales"price"of"the" sample"by"making"it"larger"and"less"representaJve"of"the"average"sales"price"of"the" majority"of"homes,"which"is"closer"to"$500,000"than"to"$600,000.""
" Preliminary"analysis"of"the"sales"data"on"the"selected"comparables"might"further"reveal"the" standard"deviaJon"for"the"sample"of"100"properJes"is"$100,000."This"means"the"sales" prices"of"the"expensive"properJes"are"more"than"2"standard"deviaJons"above"the"sample" mean"and"are"therefore"in"the"top"5%"of"sales"prices"for"the"selected"properJes.""Using"the" standard"deviaJon"to"idenJfy"staJsJcal"outliers,"we"can"exclude"them"from"the"sample" and"recalculate"the"mean"sales"price"of"the"remaining"95"properJes.""
6"