SAS code used for MODA project to preemptively identify at-risk buildings provided by Sohaib Hasan (http://blog.datalook.io/using-data-analytics-to-make-bad-buildings-better-in-new-york-city/)

PCA and models

*****PCA*****;

proc factor data=foreclosures (drop=n_dep_prty_NURSINGHOME n_dob_app_0001544 n_dob_app_0002571 n_dob_app_0037664 n_dob_app_0007320 n_dob_app_0002920) method=p priors=max rotate=promax mineigen=1.3  outstat=fact1 scree corr res score ultraheywood noprint;

var high_risk_neighborhood aep bip_score erp_charge lien_amount lis_pendens n_311: n_911: n_dep: n_dob: n_ecb: n_fires n_hpd: n_units nm_val_ttl_amt numfloors rent_stab tax_lien yearbuilt;

run;

*****Foreclosures Model*****;

proc logistic data=scores1 descending /*plots=all*/ outest=estimates;

model foreclosure = Factor1 Factor3 Factor4 Factor5 Factor6 Factor9 Factor14 Factor15 Factor18 Factor23 Factor25;

output out=foreclosures_modeled_logit predicted=p_foreclosure u=p_forc_ucl l=p_forc_lcl;

run;

*****B/C Violations Model*****;

proc qlim data=scores4 outest=hpd_bc_truncout;

model hpd_BC_per_unit = factor1 factor2 factor3 factor4 factor5 factor6 factor7 factor8 factor9  factor11 factor12 factor13 factor14  factor16

factor17 factor19  factor21 factor22 factor23 factor24 factor25  factor27;

endogenous hpd_bc_per_unit ~ censored (lb=0);

output out=BCVios_modeled_tobit predicted conditional;

run;

*****DOB Vacates Model*****

proc genmod data=scores6 /*plots=all*/;

model dobvac = factor1 factor2 factor3 factor4 factor5 factor6 factor7 factor8 factor9 factor10 factor11 factor12 factor13 factor14 factor15 factor16

factor17 factor18 factor19 factor20 factor21  factor23 factor24 factor25 factor26 factor27 factor28 / type3 dist=poisson dscale;

output out=dobvac_modeled_pois pred=p_dobvac l=p_dobvac_lcl u=p_dobvac_ucl;

run;