Page 13 - Regression Guideline

An"Example"of"SAVVI’s"Regression" Modeling"Steps"con’t..."
Step&4:&Stepwise&mulDvariable®ression&
"
In"the"fourth"step,"only"those"property"characterisJcs"significant"in"the"bivariate"models"are" retained"and"entered"into"a"single"mulJvariate"regression"model."This"model"examines"the" associaJons"among"the"remaining"variables"to"further"weed"out"those"that"overlap"with"other" variables"in"terms"of"predicJng"price."A"technique"called"stepwise"regression"(see"Important" Terms"You"Need"to"Know)"is"used"to"examine"each"variable"against"all"other"possible" variables."Only"those"variables"that"uniquely"conJnue"to"account"for"a"significant"percentage" of"the"price"variaJon"in"the"MLS"properJes"examined"remain"in"the"final"model"as"a"predictor" set."The"effects"of"the"variables"removed"from"the"model"are"subsumed"within"and"accounted" for"by"the"remaining"variables.""
"
The"overall"model"(all"remaining"property"characterisJcs"as"a"set)"is"tested"for"staJsJcal" significance"using"an"FPtest,"which"compares"the"amount"of"price"variaJon"explained"by"the" model"to"the"variaJon"that"is"not"explained."Significance"means"the"model"reliably"esJmates" price"variaJon"in"the"examined"properJes"at"the"95%"confidence"level."The"interpretaJon"of" the"coefficients,"%"price"variaJon,"and"staJsJcal"significance"is"the"same"as"in"the"bivariate" models.""
The"variables"below"are"not"staJsJcally"significant"and"do"not"uniquely"account"for"price" variaJon."They"are"excluded"from"the"final"model."Their"effect"on"price"is"accounted"for"by"the" smaller"set"of"variables"above"that"remain"staJsJcally"significant."This"makes"the"final"model" more"efficient"and"equally"accurate"as"if"these"variables"had"been"included;"in"effect,"they"are" included.""
12"