Page 28 - Regression Guideline
P. 28
Underwriters"quesJon"and"response" con’t..."
$Q.&&If&RVSquared&is&only&48%,&how&can&this®ression&analysis&be&deemed&credible?"
"
A. No"model"is"perfectly"predicJve,"regression"or"otherwise."The"quesJon"is"whether"the"
model"improves"the"ability"to"esJmate"the"sales"price"of"a"property"relaJve"to"another" method."The"48%"means"that"there"are"other"factors"that"account"for"variaJon"in"sales" price"that"were"not"included"in"the"model"and"might"not"even"be"represented"in"the"MLS." These"could"be"things"like"people"selling"homes"as"part"of"an"estate"sale"to"get"rid"of"the" home"quickly"in"which"case"the"home"might"be"undervalued"relaJve"to"what"would"be" predicted."Or"it"could"mean"someone"overvalued"a"property"and"paid"an"asking"price" that"was"too"high."Or"the"home"has"brand"new"fixtures"and"flooring"or"a"spectacular"view" that"can"not"be"modeled"using"MLS"data."These"factors"are"also"not"likely"to"be"taken" into"account"doing"an"appraisal"that"does"not"use"regression"analyses."The"regression" analysis"provides"the"best"esJmate"of"a"sales"price/property"value"using"the"known" factors"in"the"MLS."As"such,"it"improves"upon"other"methods"of"appraisal"because"it" accounts"for"the"associaJons"of"the"property"characterisJcs"with"each"other"as"well"as" sales"price"for"hundreds"of"properJes."It"improves"upon"other"methods"of"appraisal,"but" it"does"not"perfectly"predict"sales"prices.""
"
27"