Demographic Influences on Landscape Management Change in Maricopa County, AZ
28 June 2015
Michael Minn
Department of Natural Resources and Environmental Sciences
The University of Illinois Urbana-Champaign
michaelminn.com
1. Introduction
This report details the results of a preliminary investigation of the relationship between demographic characteristics, levels of vegetation, and trajectories of change in vegetation levels in census tracts in Maricopa County over the period 2002-2014.
Four different types of models were built in the statistical package R to relate NDVI and NDVI change to demographic and aggregated parcel data: linear regression models (lm), generalized additive models (gam package), random forest models (randomForest package), and conditional inference trees (partykit package)
Model fits were not particularly strong (R2 in the 300s) and different models flagged different sets of variables as significant. However, the presence of patterns in the data gives justification for further investigation.
At the macro level, additional variables, non-linear modeling approaches, and investigation of nested effects may yield fruitful results.
At the micro level, interrogation of specific high-NDVI or highly-changed NDVI tracts may help identify quantifiable characteristics that can be incorporated into the overall models.
2. Methods
2.1 Data
2.1.1 Census Data
Demographic data for 2000-2008 was taken from the 2000 census. Demographic data for 2009-2013 was taken from the last years of the American Community Survey (ACS) 5-year averages (e.g. data for 2009 taken from the 2005-2009 ACS).
2010 Census Tracts were used for spatial boundaries. Since most changes to tracts between the 2000 and 2010 census involved splitting tracts, tract values prior to 2010 were allocated to 2010 tracts contained within associated 2000 tracts.
Census variables include:
- SQMLAND: Tract area in square meters. This is a proxy for population density since census tracts usually contain around 4,000 residents
- MEDIANAGE: Median age
- MEDHHINC: Median household income
- MEANHHSIZE: Mean number of members in each household
- PCOWNEROCC: Percent housing units owner-occupied
- PCTURNOVER: Percent residents in same house one year ago
- PCBORNUSA: Percent of residents born in the USA
- PCUNEMPLOYED: Percent 16 years of age or older unemployed
2.1.2 Landsat Data
Level of vegetation in a tract was measured using the median Normalized Difference Vegetation Index (NDVI) for all Landsat pixels with centroids within the boundaries within that tract.
NDVI is a normalized ratio of reflected near infrared (NIR) light, which is related to the height and total area of vegetation, and red light, which is related to health (or "greenness") of the vegetation, and is calculated as:
NDVI = (NIR - RED) / (NIR + RED)
For this analysis we used Landsat path 37, row 37, which covered 95% of the county's residential parcels. The Landsat series of sensors captures biweekly imagery (approximately 26 images per year) at 30 m2 resolution, and have a historical archive dating back to approximately 1984 from which NDVI can be calculated. We used Landsat Five Thematic Mapper (TM) data for the period 4 January 2002 through 4 November 2011 and Landsat Seven Enhanced Thematic Mapper Plus (ETMþ) data for the period 12 January 2002 through 31 January 2012 that was downloaded from the USGS EarthExplorer website.
The fmask raster (Zhu & Woodcock, 2012) provided with the Landsat multi- spectral data permitted masking of cloud-covered areas. Some data was also missing from the Landsat Seven scenes due to the scan- line-corrector (SLC) problem, which results in a loss of around 22% of the pixels within any given scene (USGS, 2013). Because the two satellite systems provided interleaved 16-day passes, no attempt was made to interpolate missing data within individual scenes.
2.1.3 Parcel Data
Tract level data on parcels was derived from from the 2013 ST 42030 residential parcel data file and parcel shapefiles acquired from the Maricopa County Assessor's Office.
Potential lawn area for each parcel was calculated as:
PLA = parcel_area - (building_living_area / floors) - pool_area
Foreclosure data was acquired from a commercial vendor (The Information Market, 2013) covering the period from 2002-2012. The database contained 462,380 records, representing 341,983 distinct parcels or approximately 28% of the 1,214,578 residential parcels documented by the Maricopa County Assessor's Office (2013). Of the foreclosure records, 242,638 (52% of total records) indicated a completed foreclosure (Trustee's Deed or TD), representing 232,266 distinct parcels or 19% of total number of residential parcels as of 2013.
- SQMLAWN: Total square meters of PLA in each tract
- MEDCONSTYR: Median construction year
- FORECLOSED: Number of parcels in period after auction but before next sale
2.1.4 Climate Data
NDVI is affected by climate since vegetation levels rise and fall with as climate conditions became more and less favorable for growth. Rainfall alone and rainfall minus potential evapotranspiration (PET) were tried as climate signals. Running sums of 30 and 60 days were used to account for the persistence of moisture in the environment after otherwise transient rainfall events. Various temporal lags were also tested to account for both delays in vegetation response to changed conditions and sampling anomalies caused by the 16-day cycle of Landsat observations.
The following climate variables were considered:
- P_MINUS_PET: Precipitation - potential evapotranspiration
- P_MINUS_PET30: Running sum of P - PET for past 30 days
- P_MINUS_PET60: Running sum of P - PET for past 60 days
- RAINFALL30: Running sum of rainfall for the past 30 days
- RAINFALL60: Running sum of rainfall for the past 60 days
The results of correlating these variables with overall county-wide NDVI at various temporal lags resulted in the following:
- A temporally-lagged running sum of rainfall has the highest correlation with tract NDVI.
- A five-day lag between P_MINUS_PET and NDVI gives best correlation (r = 0.18)
- A nine-day lag between RAINFALL60 and NDVI gives best correlation (r = 0.423)
Accordingly, RAINFALL60 with a nine-day lag is used as a climate variable in the models below.
Correlation of climate variables with overall NDVI at various time lags
(analysis/regression-tract/climate-ndvi-correlation.png)
2.1.5 Dummy Variable
The GEOINDEX dummy variable is a sequential index of tracts in the database ordered by census bureau GEOID.
2.2 Data Organization
Two different organizational structures were used to analyze a synthesis of the demographic, parcel, and NDVI data.
Panel Analysis: The first structure involved one large table with separate rows for each NDVI date in each census tract. Columns were NDVI values and associated ACS and aggregated parcel data on the given NDVI dates. This structure permitted the creation of regression models (panel regression?) relating NDVI to demographic and parcel data across time and space.
Trend Analysis: The second structure involved performing linear regression for each census tract on the NDVI time series associated with that tract. The resulting slope coefficients indicate the trajectory of vegetation change (if any) on each census tract and the intercept coefficients indicate the overall level of vegetation relative to other census tracts. Regression models were then created to relate general vegetation level and vegetation change to 2013 ACS data and aggregated 2013 parcel data.
2.3 Models
Four different types of models were used to relate NDVI and NDVI change to demographic and aggregated parcel data: linear regression models, generalized additive models (gam package), random forest models (randomForest package), and conditional inference trees (partykit)
Variables were normalized (z-score) with the linear regression models so coefficients reflected the comparative importance of the variables to the model.
Variables with all other models were left unnormalized so the values in the model outputs could be interpreted in the context of the ranges of the individual variables.
While additional variables other than the ones listed above were investigated, variables that were strongly correlated to each other were removed to avoid confusing the GAM. Variable correlations are listed in the appendix.
3. Results
3.1 Panel Analysis
Panel analysis with the four models resulted in conflicting results about which predictors were most important, reflecting the relative weakness of all models. The ranking of importance (in decreasing order) from the linear model and random forest model:
Rank Linear Model Random Forest GAM ----------------------------------------------------- 1 MEANHHSIZE MEDCONSTYR MEDCONSTYR 2 RAINFALL60 SQMLAWN SQMLAWN 3 MEDIANAGE MEDHHINC MEDHHINC 4 MEDCONSTYR MEANHHSIZE MEDIANAGE 5 PCUNEMPLOYED MEDIANAGE GEOINDEX 6 PCBORNUSA GEOINDEX DATE 7 PCTURNOVER SQMLAND MEANHHSIZE 8 FORECLOSED PCOWNEROCC FORECLOSED 9 PCOWNEROCC PCBORNUSA RAINFALL60 10 GEOINDEX PCTURNOVER SQMLAND 11 DATE PCUNEMPLOYED PCOWNEROCC 12 MEDHHINC SQMWATER PCTURNOVER 13 SQMLAWN RAINFALL60 PCBORNUSA 14 SQMLAND FORECLOSED PCUNEMPLOYED -----------------------------------------------------
3.1.1 Linear Model
As shown in the graph below, the linear model fit was modestly strong (R2 = 0.252). As would be expected, the climate signal (RAINFALL60) had a strong positive influence on NDVI. Median year of construction (MEDCONSTYR) had a strong negative influence on NDVI, implying that neighborhoods have less vegetation the newer they are. However, the comparatively strong negative relationship between NDVI and household size (MEANHHSIZE), median age (MEDIANAGE) and percent unemployment (PCUNEMPLOYED) is less intuitive. Median household income (MEDHHINC) is found to have only a limited influence in this linear model.
Linear model NDVI prediction vs actual values
(analysis/regression-tract/prediction-linear.png)
Call:
lm(formula = NDVI ~ ., data = regression_data)
Residuals:
Min 1Q Median 3Q Max
-0.21636 -0.02755 -0.00633 0.02014 0.68708
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.923e+00 1.409e-02 207.454 <2e-16 ***
MEANHHSIZE -2.159e-02 2.482e-04 -86.988 <2e-16 ***
RAINFALL60 6.524e-03 8.389e-05 77.764 <2e-16 ***
MEDIANAGE -1.575e-03 1.492e-05 -105.527 <2e-16 ***
MEDCONSTYR -1.351e-03 7.487e-06 -180.432 <2e-16 ***
PCUNEMPLOYED -4.232e-04 4.832e-05 -8.758 <2e-16 ***
PCBORNUSA 3.597e-04 1.355e-05 26.539 <2e-16 ***
PCTURNOVER 2.455e-04 6.683e-06 36.741 <2e-16 ***
FORECLOSED -2.019e-04 8.033e-06 -25.138 <2e-16 ***
PCOWNEROCC 6.044e-06 2.676e-07 22.583 <2e-16 ***
GEOINDEX -4.399e-06 3.693e-07 -11.912 <2e-16 ***
DATE -1.275e-06 1.328e-07 -9.603 <2e-16 ***
MEDHHINC 6.273e-07 5.064e-09 123.878 <2e-16 ***
SQMLAWN 3.028e-09 7.483e-11 40.469 <2e-16 ***
SQMLAND -9.675e-11 6.079e-12 -15.917 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.04305 on 216291 degrees of freedom
(64679 observations deleted due to missingness)
Multiple R-squared: 0.2515, Adjusted R-squared: 0.2515
F-statistic: 5192 on 14 and 216291 DF, p-value: < 2.2e-16
3.1.2 Generalized Additive Model
As shown in the graph below, the generalized additive model (GAM) fit was stronger than the linear model (R2 = 0.412) and the smoothing graphs show the variable relationships to frequently be non-linear. Median construction year (MEDCONSTYR) and median age (MEDIANAGE) have strong negative relationships to NDVI while median household income (MEDHHINC) and amount of potential lawn area (SQMLAWN) have positive relationships to NDVI. Other variables are more ambiguous in their relationship to NDVI.
Generalized additive model NDVI prediction vs actual values
(analysis/regression-tract/prediction-gam.png)
Generalized additive model smoothing functions
(analysis/regression-tract/panel-gam-smooth.png)
Family: gaussian
Link function: identity
Formula:
NDVI ~ s(SQMLAND) + s(DATE) + s(MEDIANAGE) + s(MEDHHINC) + s(MEANHHSIZE) +
s(PCOWNEROCC) + s(PCTURNOVER) + s(PCBORNUSA) + s(PCUNEMPLOYED) +
s(SQMLAWN) + s(MEDCONSTYR) + s(FORECLOSED) + s(RAINFALL60) +
s(GEOINDEX)
Parametric coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.995e-01 8.203e-05 2432 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Approximate significance of smooth terms:
edf Ref.df F p-value
s(MEDCONSTYR) 8.983 9.000 3043.24 <2e-16 ***
s(SQMLAWN) 8.983 9.000 1413.16 <2e-16 ***
s(MEDHHINC) 8.999 9.000 1397.03 <2e-16 ***
s(MEDIANAGE) 8.998 9.000 1157.59 <2e-16 ***
s(GEOINDEX) 8.992 9.000 1098.53 <2e-16 ***
s(DATE) 8.992 9.000 942.75 <2e-16 ***
s(MEANHHSIZE) 8.986 9.000 835.43 <2e-16 ***
s(FORECLOSED) 8.005 8.474 695.20 <2e-16 ***
s(RAINFALL60) 8.993 9.000 570.55 <2e-16 ***
s(SQMLAND) 8.976 9.000 365.73 <2e-16 ***
s(PCOWNEROCC) 8.986 9.000 323.47 <2e-16 ***
s(PCTURNOVER) 8.984 9.000 209.19 <2e-16 ***
s(PCBORNUSA) 8.964 9.000 186.42 <2e-16 ***
s(PCUNEMPLOYED) 8.931 8.997 34.45 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
R-sq.(adj) = 0.412 Deviance explained = 41.2%
GCV = 0.0014562 Scale est. = 0.0014554 n = 216306
3.1.3 Random Forest Model
As shown below, the random forest model found median construction year (MEDCONSTYR) to be highly important, with amount of PLA (SQMLAWN), median household income (MEDHHINC), median household size (MEDHHSIZE) and median age (MEDIANAGE) to also be important, which is consistent with the linear and GAM results.
The random forest model also found minimal interaction between variables.
Random forest model NDVI prediction vs actual values
(analysis/regression-tract/prediction-rf.png)
> plot(randmodel)
Importance Relative Imp
MEDCONSTYR 0.0031 1.0000
SQMLAWN 0.0015 0.4960
MEDHHINC 0.0013 0.4176
MEANHHSIZE 0.0013 0.4149
MEDIANAGE 0.0010 0.3285
GEOINDEX 0.0010 0.3085
SQMLAND 0.0009 0.2967
PCOWNEROCC 0.0007 0.2268
PCBORNUSA 0.0006 0.1913
PCTURNOVER 0.0006 0.1874
PCUNEMPLOYED 0.0004 0.1169
SQMWATER 0.0004 0.1151
RAINFALL60 0.0002 0.0760
FORECLOSED 0.0002 0.0579
3.1.4 Conditional Inference Tree (Party)
The conditional inference tree was the most ambiguous of the four models, evidencing no unambiguously strong predictors.
Conditional Inference Tree model NDVI prediction vs actual values
(analysis/regression-tract/prediction-party.png)
> print(party)
Model formula:
NDVI ~ DATE + SQMLAND + MEDIANAGE + MEDHHINC + MEANHHSIZE + PCOWNEROCC +
PCTURNOVER + PCBORNUSA + PCUNEMPLOYED + SQMLAWN + MEDCONSTYR +
FORECLOSED + RAINFALL60 + GEOINDEX
Fitted party:
[1] root
| [2] SQMLAND <= 269339.64455
| | [3] SQMLAND <= 227883.48089
| | | [4] MEDCONSTYR <= 1969
| | | | [5] PCTURNOVER <= 45.2
| | | | | [6] MEDHHINC <= 30447.33256
| | | | | | [7] GEOINDEX <= 651
| | | | | | | [8] PCBORNUSA <= 60.6
| | | | | | | | [9] MEANHHSIZE <= 4.05: 0.205 (n = 519, err = 0.4)
| | | | | | | | [10] MEANHHSIZE > 4.05: 0.171 (n = 143, err = 0.1)
| | | | | | | [11] PCBORNUSA > 60.6
| | | | | | | | [12] GEOINDEX <= 149: 0.274 (n = 42, err = 0.0)
| | | | | | | | [13] GEOINDEX > 149: 0.227 (n = 168, err = 0.1)
| | | | | | [14] GEOINDEX > 651
| | | | | | | [15] RAINFALL60 <= 1.34
| | | | | | | | [16] PCOWNEROCC <= 26.7: 0.181 (n = 253, err = 0.2)
| | | | | | | | [17] PCOWNEROCC > 26.7: 0.157 (n = 282, err = 0.1)
| | | | | | | [18] RAINFALL60 > 1.34
| | | | | | | | [19] RAINFALL60 <= 3.55: 0.183 (n = 193, err = 0.1)
| | | | | | | | [20] RAINFALL60 > 3.55: 0.202 (n = 69, err = 0.1)
| | | | | [21] MEDHHINC > 30447.33256
| | | | | | [22] PCBORNUSA <= 91
| | | | | | | [23] MEDIANAGE <= 26.4
| | | | | | | | [24] PCOWNEROCC <= 46.7: 0.218 (n = 1869, err = 2.0)
| | | | | | | | [25] PCOWNEROCC > 46.7: 0.195 (n = 777, err = 0.4)
| | | | | | | [26] MEDIANAGE > 26.4
| | | | | | | | [27] MEDHHINC <= 44754.25779: 0.254 (n = 4752, err = 13.1)
| | | | | | | | [28] MEDHHINC > 44754.25779: 0.220 (n = 910, err = 1.5)
| | | | | | [29] PCBORNUSA > 91
| | | | | | | [30] MEDHHINC <= 47714.75143: 0.282 (n = 144, err = 0.1)
| | | | | | | [31] MEDHHINC > 47714.75143
| | | | | | | | [32] GEOINDEX <= 263: 0.297 (n = 53, err = 0.1)
| | | | | | | | [33] GEOINDEX > 263: 0.315 (n = 128, err = 0.1)
| | | | [34] PCTURNOVER > 45.2
| | | | | [35] MEANHHSIZE <= 2.48
| | | | | | [36] PCTURNOVER <= 85.5
| | | | | | | [37] SQMLAWN <= 437819.4623
| | | | | | | | [38] PCTURNOVER <= 56.2: 0.186 (n = 528, err = 0.3)
| | | | | | | | [39] PCTURNOVER > 56.2: 0.234 (n = 2227, err = 5.9)
| | | | | | | [40] SQMLAWN > 437819.4623
| | | | | | | | [41] MEDIANAGE <= 44.6: 0.297 (n = 1783, err = 3.1)
| | | | | | | | [42] MEDIANAGE > 44.6: 0.221 (n = 244, err = 0.1)
| | | | | | [43] PCTURNOVER > 85.5
| | | | | | | [44] PCBORNUSA <= 89.2
| | | | | | | | [45] MEANHHSIZE <= 2.06: 0.323 (n = 87, err = 0.2)
| | | | | | | | [46] MEANHHSIZE > 2.06: 0.231 (n = 204, err = 0.3)
| | | | | | | [47] PCBORNUSA > 89.2
| | | | | | | | [48] GEOINDEX <= 537: 0.332 (n = 397, err = 0.7)
| | | | | | | | [49] GEOINDEX > 537: 0.258 (n = 64, err = 0.1)
| | | | | [50] MEANHHSIZE > 2.48
| | | | | | [51] PCUNEMPLOYED <= 1
| | | | | | | [52] SQMLAND <= 121309.19091
| | | | | | | | [53] MEDHHINC <= 43381.97655: 0.193 (n = 760, err = 1.1)
| | | | | | | | [54] MEDHHINC > 43381.97655: 0.222 (n = 372, err = 0.6)
| | | | | | | [55] SQMLAND > 121309.19091
| | | | | | | | [56] MEANHHSIZE <= 3.18: 0.269 (n = 236, err = 0.4)
| | | | | | | | [57] MEANHHSIZE > 3.18: 0.198 (n = 205, err = 0.3)
| | | | | | [58] PCUNEMPLOYED > 1
| | | | | | | [59] SQMLAWN <= 232665.57643
| | | | | | | | [60] MEDHHINC <= 25094.87004: 0.150 (n = 953, err = 0.6)
| | | | | | | | [61] MEDHHINC > 25094.87004: 0.187 (n = 981, err = 1.1)
| | | | | | | [62] SQMLAWN > 232665.57643
| | | | | | | | [63] MEDCONSTYR <= 1955: 0.214 (n = 5299, err = 10.6)
| | | | | | | | [64] MEDCONSTYR > 1955: 0.191 (n = 7877, err = 8.8)
| | | [65] MEDCONSTYR > 1969
| | | | [66] FORECLOSED <= 29
| | | | | [67] MEDCONSTYR <= 1993
| | | | | | [68] MEANHHSIZE <= 2.96
| | | | | | | [69] RAINFALL60 <= 0.48
| | | | | | | | [70] MEDCONSTYR <= 1974: 0.206 (n = 2389, err = 3.8)
| | | | | | | | [71] MEDCONSTYR > 1974: 0.192 (n = 10227, err = 11.0)
| | | | | | | [72] RAINFALL60 > 0.48
| | | | | | | | [73] PCTURNOVER <= 43.9: 0.212 (n = 5028, err = 7.8)
| | | | | | | | [74] PCTURNOVER > 43.9: 0.200 (n = 11983, err = 14.9)
| | | | | | [75] MEANHHSIZE > 2.96
| | | | | | | [76] PCBORNUSA <= 77.9
| | | | | | | | [77] RAINFALL60 <= 2.35: 0.174 (n = 6233, err = 4.7)
| | | | | | | | [78] RAINFALL60 > 2.35: 0.189 (n = 1273, err = 1.5)
| | | | | | | [79] PCBORNUSA > 77.9
| | | | | | | | [80] RAINFALL60 <= 1.84: 0.188 (n = 4023, err = 3.8)
| | | | | | | | [81] RAINFALL60 > 1.84: 0.203 (n = 1068, err = 1.3)
| | | | | [82] MEDCONSTYR > 1993
| | | | | | [83] MEDHHINC <= 75723.69581
| | | | | | | [84] PCBORNUSA <= 94.6
| | | | | | | | [85] PCUNEMPLOYED <= 1.9: 0.147 (n = 3101, err = 3.4)
| | | | | | | | [86] PCUNEMPLOYED > 1.9: 0.170 (n = 8701, err = 9.6)
| | | | | | | [87] PCBORNUSA > 94.6
| | | | | | | | [88] MEANHHSIZE <= 1.2: 0.077 (n = 56, err = 0.0)
| | | | | | | | [89] MEANHHSIZE > 1.2: 0.144 (n = 1285, err = 0.7)
| | | | | | [90] MEDHHINC > 75723.69581
| | | | | | | [91] PCOWNEROCC <= 633
| | | | | | | | [92] MEDHHINC <= 124261.10286: 0.180 (n = 2859, err = 2.3)
| | | | | | | | [93] MEDHHINC > 124261.10286: 0.249 (n = 82, err = 0.0)
| | | | | | | [94] PCOWNEROCC > 633
| | | | | | | | [95] PCOWNEROCC <= 955: 0.211 (n = 522, err = 0.5)
| | | | | | | | [96] PCOWNEROCC > 955: 0.182 (n = 267, err = 0.1)
| | | | [97] FORECLOSED > 29
| | | | | [98] MEANHHSIZE <= 2.81
| | | | | | [99] MEDIANAGE <= 38.1
| | | | | | | [100] PCTURNOVER <= 61.8
| | | | | | | | [101] MEDHHINC <= 46993.05573: 0.186 (n = 107, err = 0.0)
| | | | | | | | [102] MEDHHINC > 46993.05573: 0.168 (n = 65, err = 0.0)
| | | | | | | [103] PCTURNOVER > 61.8
| | | | | | | | [104] DATE <= 14994: 0.207 (n = 461, err = 0.4)
| | | | | | | | [105] DATE > 14994: 0.182 (n = 96, err = 0.1)
| | | | | | [106] MEDIANAGE > 38.1
| | | | | | | [107] MEDCONSTYR <= 2001
| | | | | | | | [108] FORECLOSED <= 36: 0.179 (n = 178, err = 0.1)
| | | | | | | | [109] FORECLOSED > 36: 0.167 (n = 82, err = 0.0)
| | | | | | | [110] MEDCONSTYR > 2001: 0.138 (n = 33, err = 0.0)
| | | | | [111] MEANHHSIZE > 2.81
| | | | | | [112] SQMLAWN <= 935428.0438
| | | | | | | [113] PCOWNEROCC <= 786
| | | | | | | | [114] MEDCONSTYR <= 1992: 0.176 (n = 1684, err = 0.7)
| | | | | | | | [115] MEDCONSTYR > 1992: 0.155 (n = 1355, err = 0.6)
| | | | | | | [116] PCOWNEROCC > 786
| | | | | | | | [117] RAINFALL60 <= 0.67: 0.172 (n = 276, err = 0.1)
| | | | | | | | [118] RAINFALL60 > 0.67: 0.193 (n = 368, err = 0.2)
| | | | | | [119] SQMLAWN > 935428.0438: 0.204 (n = 165, err = 0.1)
| | [120] SQMLAND > 227883.48089
| | | [121] PCTURNOVER <= 21
| | | | [122] MEDIANAGE <= 28.9
| | | | | [123] PCBORNUSA <= 93.8
| | | | | | [124] SQMLAWN <= 664295.65507
| | | | | | | [125] DATE <= 12746: 0.233 (n = 213, err = 0.3)
| | | | | | | [126] DATE > 12746
| | | | | | | | [127] PCTURNOVER <= 4: 0.194 (n = 38, err = 0.0)
| | | | | | | | [128] PCTURNOVER > 4: 0.218 (n = 103, err = 0.1)
| | | | | | [129] SQMLAWN > 664295.65507
| | | | | | | [130] MEDCONSTYR <= 1999
| | | | | | | | [131] SQMLAWN <= 997626.27714: 0.192 (n = 123, err = 0.0)
| | | | | | | | [132] SQMLAWN > 997626.27714: 0.211 (n = 399, err = 0.2)
| | | | | | | [133] MEDCONSTYR > 1999
| | | | | | | | [134] FORECLOSED <= 6: 0.164 (n = 21, err = 0.0)
| | | | | | | | [135] FORECLOSED > 6: 0.189 (n = 44, err = 0.0)
| | | | | [136] PCBORNUSA > 93.8
| | | | | | [137] PCBORNUSA <= 93.9
| | | | | | | [138] DATE <= 12498: 0.155 (n = 67, err = 0.0)
| | | | | | | [139] DATE > 12498
| | | | | | | | [140] FORECLOSED <= 18: 0.174 (n = 131, err = 0.0)
| | | | | | | | [141] FORECLOSED > 18: 0.193 (n = 32, err = 0.0)
| | | | | | [142] PCBORNUSA > 93.9
| | | | | | | [143] FORECLOSED <= 23: 0.127 (n = 43, err = 0.0)
| | | | | | | [144] FORECLOSED > 23: 0.144 (n = 48, err = 0.0)
| | | | [145] MEDIANAGE > 28.9
| | | | | [146] PCBORNUSA <= 89.9
| | | | | | [147] PCUNEMPLOYED <= 2.8
| | | | | | | [148] PCUNEMPLOYED <= 1.2
| | | | | | | | [149] DATE <= 12498: 0.139 (n = 57, err = 0.0)
| | | | | | | | [150] DATE > 12498: 0.161 (n = 158, err = 0.1)
| | | | | | | [151] PCUNEMPLOYED > 1.2
| | | | | | | | [152] MEDHHINC <= 55764.68036: 0.153 (n = 225, err = 0.0)
| | | | | | | | [153] MEDHHINC > 55764.68036: 0.137 (n = 878, err = 0.6)
| | | | | | [154] PCUNEMPLOYED > 2.8
| | | | | | | [155] RAINFALL60 <= 1.34
| | | | | | | | [156] FORECLOSED <= 3: 0.198 (n = 106, err = 0.0)
| | | | | | | | [157] FORECLOSED > 3: 0.209 (n = 31, err = 0.0)
| | | | | | | [158] RAINFALL60 > 1.34: 0.211 (n = 62, err = 0.0)
| | | | | [159] PCBORNUSA > 89.9
| | | | | | [160] SQMLAWN <= 716930.77875
| | | | | | | [161] PCTURNOVER <= 17.2
| | | | | | | | [162] FORECLOSED <= 1: 0.121 (n = 39, err = 0.0)
| | | | | | | | [163] FORECLOSED > 1: 0.132 (n = 22, err = 0.0)
| | | | | | | [164] PCTURNOVER > 17.2
| | | | | | | | [165] RAINFALL60 <= 3.63: 0.153 (n = 209, err = 0.0)
| | | | | | | | [166] RAINFALL60 > 3.63: 0.172 (n = 12, err = 0.0)
| | | | | | [167] SQMLAWN > 716930.77875
| | | | | | | [168] MEDCONSTYR <= 1999
| | | | | | | | [169] GEOINDEX <= 158: 0.238 (n = 197, err = 0.1)
| | | | | | | | [170] GEOINDEX > 158: 0.194 (n = 706, err = 0.8)
| | | | | | | [171] MEDCONSTYR > 1999
| | | | | | | | [172] SQMLAWN <= 1048909.80153: 0.127 (n = 167, err = 0.1)
| | | | | | | | [173] SQMLAWN > 1048909.80153: 0.162 (n = 90, err = 0.0)
| | | [174] PCTURNOVER > 21
| | | | [175] MEANHHSIZE <= 3.48
| | | | | [176] MEANHHSIZE <= 1.78
| | | | | | [177] MEDHHINC <= 60109.07165
| | | | | | | [178] SQMLAND <= 240233.09395
| | | | | | | | [179] PCBORNUSA <= 88.7: 0.148 (n = 310, err = 0.2)
| | | | | | | | [180] PCBORNUSA > 88.7: 0.137 (n = 418, err = 0.2)
| | | | | | | [181] SQMLAND > 240233.09395
| | | | | | | | [182] MEDIANAGE <= 60.7: 0.221 (n = 296, err = 0.4)
| | | | | | | | [183] MEDIANAGE > 60.7: 0.164 (n = 1457, err = 0.5)
| | | | | | [184] MEDHHINC > 60109.07165
| | | | | | | [185] MEDHHINC <= 72362.51403
| | | | | | | | [186] MEDIANAGE <= 60.2: 0.242 (n = 487, err = 0.5)
| | | | | | | | [187] MEDIANAGE > 60.2: 0.162 (n = 99, err = 0.1)
| | | | | | | [188] MEDHHINC > 72362.51403: 0.297 (n = 86, err = 0.1)
| | | | | [189] MEANHHSIZE > 1.78
| | | | | | [190] MEDIANAGE <= 37.5
| | | | | | | [191] SQMLAND <= 252155.0739
| | | | | | | | [192] MEDIANAGE <= 32.9: 0.213 (n = 36066, err = 71.8)
| | | | | | | | [193] MEDIANAGE > 32.9: 0.205 (n = 19500, err = 35.7)
| | | | | | | [194] SQMLAND > 252155.0739
| | | | | | | | [195] MEDHHINC <= 14646.41031: 0.150 (n = 303, err = 0.1)
| | | | | | | | [196] MEDHHINC > 14646.41031: 0.192 (n = 3627, err = 3.3)
| | | | | | [197] MEDIANAGE > 37.5
| | | | | | | [198] MEDCONSTYR <= 1965
| | | | | | | | [199] PCBORNUSA <= 91.3: 0.266 (n = 2385, err = 9.4)
| | | | | | | | [200] PCBORNUSA > 91.3: 0.342 (n = 1540, err = 2.9)
| | | | | | | [201] MEDCONSTYR > 1965
| | | | | | | | [202] MEDHHINC <= 66685.14033: 0.197 (n = 7935, err = 16.4)
| | | | | | | | [203] MEDHHINC > 66685.14033: 0.227 (n = 9542, err = 20.9)
| | | | [204] MEANHHSIZE > 3.48
| | | | | [205] MEANHHSIZE <= 4.03
| | | | | | [206] MEDHHINC <= 78362.10213
| | | | | | | [207] MEANHHSIZE <= 3.94
| | | | | | | | [208] SQMLAWN <= 661231.71413: 0.161 (n = 1544, err = 1.1)
| | | | | | | | [209] SQMLAWN > 661231.71413: 0.189 (n = 2314, err = 2.8)
| | | | | | | [210] MEANHHSIZE > 3.94
| | | | | | | | [211] GEOINDEX <= 780: 0.185 (n = 731, err = 0.6)
| | | | | | | | [212] GEOINDEX > 780: 0.238 (n = 337, err = 0.5)
| | | | | | [213] MEDHHINC > 78362.10213
| | | | | | | [214] PCUNEMPLOYED <= 4.1
| | | | | | | | [215] SQMLAND <= 254243.72207: 0.257 (n = 473, err = 0.8)
| | | | | | | | [216] SQMLAND > 254243.72207: 0.199 (n = 151, err = 0.1)
| | | | | | | [217] PCUNEMPLOYED > 4.1
| | | | | | | | [218] MEDCONSTYR <= 1998: 0.255 (n = 66, err = 0.1)
| | | | | | | | [219] MEDCONSTYR > 1998: 0.202 (n = 429, err = 0.3)
| | | | | [220] MEANHHSIZE > 4.03
| | | | | | [221] DATE <= 14930
| | | | | | | [222] RAINFALL60 <= 3.03
| | | | | | | | [223] SQMLAWN <= 685072.99295: 0.173 (n = 1431, err = 1.2)
| | | | | | | | [224] SQMLAWN > 685072.99295: 0.188 (n = 940, err = 0.7)
| | | | | | | [225] RAINFALL60 > 3.03
| | | | | | | | [226] RAINFALL60 <= 3.8: 0.203 (n = 216, err = 0.2)
| | | | | | | | [227] RAINFALL60 > 3.8: 0.234 (n = 49, err = 0.1)
| | | | | | [228] DATE > 14930
| | | | | | | [229] MEDIANAGE <= 21.6
| | | | | | | | [230] DATE <= 16314: 0.135 (n = 170, err = 0.1)
| | | | | | | | [231] DATE > 16314: 0.173 (n = 16, err = 0.0)
| | | | | | | [232] MEDIANAGE > 21.6
| | | | | | | | [233] SQMLAND <= 230394.46669: 0.171 (n = 58, err = 0.0)
| | | | | | | | [234] SQMLAND > 230394.46669: 0.160 (n = 335, err = 0.2)
| [235] SQMLAND > 269339.64455
| | [236] DATE <= 14090
| | | [237] SQMLAWN <= 2087419.29801
| | | | [238] SQMLAWN <= 899092.28696
| | | | | [239] SQMLAWN <= 781757.23796
| | | | | | [240] PCTURNOVER <= 35.2
| | | | | | | [241] PCOWNEROCC <= 70
| | | | | | | | [242] RAINFALL60 <= 3.12: 0.179 (n = 1592, err = 3.2)
| | | | | | | | [243] RAINFALL60 > 3.12: 0.214 (n = 134, err = 0.4)
| | | | | | | [244] PCOWNEROCC > 70
| | | | | | | | [245] MEANHHSIZE <= 3.19: 0.151 (n = 2722, err = 4.0)
| | | | | | | | [246] MEANHHSIZE > 3.19: 0.175 (n = 541, err = 1.9)
| | | | | | [247] PCTURNOVER > 35.2
| | | | | | | [248] SQMLAWN <= 748034.00637
| | | | | | | | [249] MEDCONSTYR <= 1986: 0.200 (n = 2413, err = 4.4)
| | | | | | | | [250] MEDCONSTYR > 1986: 0.170 (n = 1333, err = 2.6)
| | | | | | | [251] SQMLAWN > 748034.00637
| | | | | | | | [252] MEANHHSIZE <= 1.7: 0.307 (n = 113, err = 0.1)
| | | | | | | | [253] MEANHHSIZE > 1.7: 0.177 (n = 47, err = 0.1)
| | | | | [254] SQMLAWN > 781757.23796
| | | | | | [255] PCBORNUSA <= 92.2
| | | | | | | [256] MEDCONSTYR <= 2004
| | | | | | | | [257] PCOWNEROCC <= 34.2: 0.235 (n = 33, err = 0.3)
| | | | | | | | [258] PCOWNEROCC > 34.2: 0.167 (n = 1830, err = 1.5)
| | | | | | | [259] MEDCONSTYR > 2004
| | | | | | | | [260] DATE <= 13818: 0.100 (n = 47, err = 0.0)
| | | | | | | | [261] DATE > 13818: 0.129 (n = 28, err = 0.0)
| | | | | | [262] PCBORNUSA > 92.2
| | | | | | | [263] MEANHHSIZE <= 2.79
| | | | | | | | [264] MEANHHSIZE <= 1.84: 0.151 (n = 10, err = 0.0)
| | | | | | | | [265] MEANHHSIZE > 1.84: 0.136 (n = 319, err = 0.1)
| | | | | | | [266] MEANHHSIZE > 2.79: 0.125 (n = 92, err = 0.0)
| | | | [267] SQMLAWN > 899092.28696
| | | | | [268] MEANHHSIZE <= 3.2
| | | | | | [269] MEDHHINC <= 79904.76387
| | | | | | | [270] PCBORNUSA <= 87.5
| | | | | | | | [271] MEDIANAGE <= 34.9: 0.191 (n = 1982, err = 7.3)
| | | | | | | | [272] MEDIANAGE > 34.9: 0.309 (n = 163, err = 0.1)
| | | | | | | [273] PCBORNUSA > 87.5
| | | | | | | | [274] DATE <= 13162: 0.174 (n = 2644, err = 3.7)
| | | | | | | | [275] DATE > 13162: 0.156 (n = 3388, err = 3.5)
| | | | | | [276] MEDHHINC > 79904.76387
| | | | | | | [277] PCTURNOVER <= 35
| | | | | | | | [278] MEDCONSTYR <= 1998: 0.199 (n = 1243, err = 1.3)
| | | | | | | | [279] MEDCONSTYR > 1998: 0.155 (n = 732, err = 0.7)
| | | | | | | [280] PCTURNOVER > 35
| | | | | | | | [281] MEANHHSIZE <= 2.31: 0.249 (n = 554, err = 2.2)
| | | | | | | | [282] MEANHHSIZE > 2.31: 0.199 (n = 1004, err = 1.1)
| | | | | [283] MEANHHSIZE > 3.2
| | | | | | [284] FORECLOSED <= 14
| | | | | | | [285] GEOINDEX <= 414
| | | | | | | | [286] MEDCONSTYR <= 1996: 0.274 (n = 419, err = 1.5)
| | | | | | | | [287] MEDCONSTYR > 1996: 0.210 (n = 525, err = 1.3)
| | | | | | | [288] GEOINDEX > 414
| | | | | | | | [289] MEDHHINC <= 70369.07143: 0.180 (n = 587, err = 0.5)
| | | | | | | | [290] MEDHHINC > 70369.07143: 0.216 (n = 417, err = 0.4)
| | | | | | [291] FORECLOSED > 14
| | | | | | | [292] SQMLAWN <= 1237803.75789
| | | | | | | | [293] PCTURNOVER <= 59.1: 0.184 (n = 98, err = 0.1)
| | | | | | | | [294] PCTURNOVER > 59.1: 0.147 (n = 46, err = 0.0)
| | | | | | | [295] SQMLAWN > 1237803.75789
| | | | | | | | [296] MEANHHSIZE <= 3.29: 0.207 (n = 36, err = 0.0)
| | | | | | | | [297] MEANHHSIZE > 3.29: 0.266 (n = 9, err = 0.0)
| | | [298] SQMLAWN > 2087419.29801
| | | | [299] PCTURNOVER <= 52.6
| | | | | [300] RAINFALL60 <= 3.12
| | | | | | [301] GEOINDEX <= 551
| | | | | | | [302] MEDIANAGE <= 33.5
| | | | | | | | [303] PCBORNUSA <= 93.3: 0.224 (n = 623, err = 1.2)
| | | | | | | | [304] PCBORNUSA > 93.3: 0.296 (n = 402, err = 1.0)
| | | | | | | [305] MEDIANAGE > 33.5
| | | | | | | | [306] SQMLAWN <= 6039089.74485: 0.175 (n = 1569, err = 1.7)
| | | | | | | | [307] SQMLAWN > 6039089.74485: 0.199 (n = 523, err = 0.7)
| | | | | | [308] GEOINDEX > 551
| | | | | | | [309] MEDCONSTYR <= 2001
| | | | | | | | [310] RAINFALL60 <= 1.26: 0.177 (n = 2851, err = 3.2)
| | | | | | | | [311] RAINFALL60 > 1.26: 0.202 (n = 717, err = 1.2)
| | | | | | | [312] MEDCONSTYR > 2001
| | | | | | | | [313] SQMLAWN <= 2715552.59968: 0.135 (n = 253, err = 0.2)
| | | | | | | | [314] SQMLAWN > 2715552.59968: 0.175 (n = 345, err = 0.5)
| | | | | [315] RAINFALL60 > 3.12
| | | | | | [316] RAINFALL60 <= 3.8
| | | | | | | [317] SQMLAWN <= 6335555.6204
| | | | | | | | [318] MEDHHINC <= 93089.99896: 0.241 (n = 185, err = 0.8)
| | | | | | | | [319] MEDHHINC > 93089.99896: 0.211 (n = 128, err = 0.1)
| | | | | | | [320] SQMLAWN > 6335555.6204
| | | | | | | | [321] PCUNEMPLOYED <= 2.4: 0.267 (n = 105, err = 0.5)
| | | | | | | | [322] PCUNEMPLOYED > 2.4: 0.356 (n = 7, err = 0.1)
| | | | | | [323] RAINFALL60 > 3.8
| | | | | | | [324] DATE <= 11706: 0.192 (n = 26, err = 0.1)
| | | | | | | [325] DATE > 11706
| | | | | | | | [326] MEDHHINC <= 55783.87465: 0.388 (n = 34, err = 0.1)
| | | | | | | | [327] MEDHHINC > 55783.87465: 0.288 (n = 90, err = 0.3)
| | | | [328] PCTURNOVER > 52.6
| | | | | [329] MEDIANAGE <= 50.7
| | | | | | [330] MEDHHINC <= 120089.94422
| | | | | | | [331] PCBORNUSA <= 89.2
| | | | | | | | [332] MEDCONSTYR <= 1984: 0.231 (n = 125, err = 0.1)
| | | | | | | | [333] MEDCONSTYR > 1984: 0.205 (n = 170, err = 0.3)
| | | | | | | [334] PCBORNUSA > 89.2
| | | | | | | | [335] PCTURNOVER <= 55.3: 0.266 (n = 381, err = 0.3)
| | | | | | | | [336] PCTURNOVER > 55.3: 0.243 (n = 470, err = 0.5)
| | | | | | [337] MEDHHINC > 120089.94422
| | | | | | | [338] SQMLAND <= 351299.71479: 0.381 (n = 14, err = 0.0)
| | | | | | | [339] SQMLAND > 351299.71479
| | | | | | | | [340] MEDHHINC <= 189772.20939: 0.286 (n = 431, err = 0.3)
| | | | | | | | [341] MEDHHINC > 189772.20939: 0.259 (n = 165, err = 0.1)
| | | | | [342] MEDIANAGE > 50.7
| | | | | | [343] SQMLAWN <= 2316452.46004
| | | | | | | [344] SQMLAWN <= 2315708.0283: 0.171 (n = 9, err = 0.0)
| | | | | | | [345] SQMLAWN > 2315708.0283: 0.153 (n = 99, err = 0.0)
| | | | | | [346] SQMLAWN > 2316452.46004
| | | | | | | [347] FORECLOSED <= 1: 0.146 (n = 39, err = 0.0)
| | | | | | | [348] FORECLOSED > 1: 0.136 (n = 19, err = 0.0)
| | [349] DATE > 14090
| | | [350] MEANHHSIZE <= 1.88
| | | | [351] PCTURNOVER <= 84.7
| | | | | [352] MEDHHINC <= 75812.06149
| | | | | | [353] PCBORNUSA <= 92.6
| | | | | | | [354] MEDHHINC <= 68845.34233
| | | | | | | | [355] PCUNEMPLOYED <= 5.5: 0.186 (n = 420, err = 0.3)
| | | | | | | | [356] PCUNEMPLOYED > 5.5: 0.210 (n = 129, err = 0.1)
| | | | | | | [357] MEDHHINC > 68845.34233
| | | | | | | | [358] GEOINDEX <= 243: 0.225 (n = 39, err = 0.0)
| | | | | | | | [359] GEOINDEX > 243: 0.223 (n = 94, err = 0.1)
| | | | | | [360] PCBORNUSA > 92.6
| | | | | | | [361] SQMLAWN <= 1315349.38103
| | | | | | | | [362] DATE <= 14858: 0.177 (n = 156, err = 0.1)
| | | | | | | | [363] DATE > 14858: 0.192 (n = 118, err = 0.1)
| | | | | | | [364] SQMLAWN > 1315349.38103
| | | | | | | | [365] SQMLAND <= 480232.25782: 0.168 (n = 255, err = 0.1)
| | | | | | | | [366] SQMLAND > 480232.25782: 0.154 (n = 258, err = 0.1)
| | | | | [367] MEDHHINC > 75812.06149
| | | | | | [368] SQMLAWN <= 780206.12947: 0.274 (n = 19, err = 0.2)
| | | | | | [369] SQMLAWN > 780206.12947: 0.300 (n = 170, err = 0.1)
| | | | [370] PCTURNOVER > 84.7
| | | | | [371] PCUNEMPLOYED <= 3.9
| | | | | | [372] FORECLOSED <= 15
| | | | | | | [373] DATE <= 14994
| | | | | | | | [374] MEDCONSTYR <= 1993: 0.171 (n = 552, err = 0.2)
| | | | | | | | [375] MEDCONSTYR > 1993: 0.162 (n = 297, err = 0.1)
| | | | | | | [376] DATE > 14994
| | | | | | | | [377] PCUNEMPLOYED <= 1.7: 0.148 (n = 688, err = 0.2)
| | | | | | | | [378] PCUNEMPLOYED > 1.7: 0.157 (n = 923, err = 0.5)
| | | | | | [379] FORECLOSED > 15
| | | | | | | [380] MEDHHINC <= 48629.15449
| | | | | | | | [381] MEDCONSTYR <= 1962: 0.206 (n = 38, err = 0.0)
| | | | | | | | [382] MEDCONSTYR > 1962: 0.161 (n = 136, err = 0.0)
| | | | | | | [383] MEDHHINC > 48629.15449: 0.216 (n = 50, err = 0.0)
| | | | | [384] PCUNEMPLOYED > 3.9
| | | | | | [385] PCBORNUSA <= 92.8
| | | | | | | [386] MEDCONSTYR <= 1972: 0.183 (n = 20, err = 0.0)
| | | | | | | [387] MEDCONSTYR > 1972: 0.204 (n = 117, err = 0.1)
| | | | | | [388] PCBORNUSA > 92.8
| | | | | | | [389] DATE <= 14602
| | | | | | | | [390] MEANHHSIZE <= 1.56: 0.190 (n = 19, err = 0.0)
| | | | | | | | [391] MEANHHSIZE > 1.56: 0.162 (n = 212, err = 0.1)
| | | | | | | [392] DATE > 14602: 0.198 (n = 39, err = 0.0)
| | | [393] MEANHHSIZE > 1.88
| | | | [394] MEDHHINC <= 77947.25845
| | | | | [395] FORECLOSED <= 34
| | | | | | [396] RAINFALL60 <= 2.23
| | | | | | | [397] DATE <= 14978
| | | | | | | | [398] MEDIANAGE <= 49: 0.198 (n = 5223, err = 12.3)
| | | | | | | | [399] MEDIANAGE > 49: 0.168 (n = 626, err = 0.7)
| | | | | | | [400] DATE > 14978
| | | | | | | | [401] FORECLOSED <= 6: 0.186 (n = 3812, err = 8.9)
| | | | | | | | [402] FORECLOSED > 6: 0.177 (n = 4213, err = 7.4)
| | | | | | [403] RAINFALL60 > 2.23
| | | | | | | [404] DATE <= 14882
| | | | | | | | [405] PCTURNOVER <= 59.1: 0.191 (n = 791, err = 1.9)
| | | | | | | | [406] PCTURNOVER > 59.1: 0.218 (n = 886, err = 2.5)
| | | | | | | [407] DATE > 14882
| | | | | | | | [408] DATE <= 16330: 0.180 (n = 1947, err = 3.8)
| | | | | | | | [409] DATE > 16330: 0.208 (n = 622, err = 1.2)
| | | | | [410] FORECLOSED > 34
| | | | | | [411] GEOINDEX <= 526
| | | | | | | [412] SQMLAND <= 3381973.63409
| | | | | | | | [413] PCBORNUSA <= 89.1: 0.179 (n = 2287, err = 2.7)
| | | | | | | | [414] PCBORNUSA > 89.1: 0.165 (n = 698, err = 0.4)
| | | | | | | [415] SQMLAND > 3381973.63409
| | | | | | | | [416] GEOINDEX <= 220: 0.197 (n = 173, err = 0.3)
| | | | | | | | [417] GEOINDEX > 220: 0.264 (n = 72, err = 0.1)
| | | | | | [418] GEOINDEX > 526
| | | | | | | [419] SQMLAWN <= 2857198.69153
| | | | | | | | [420] MEDCONSTYR <= 2003: 0.164 (n = 485, err = 0.3)
| | | | | | | | [421] MEDCONSTYR > 2003: 0.133 (n = 455, err = 0.3)
| | | | | | | [422] SQMLAWN > 2857198.69153
| | | | | | | | [423] RAINFALL60 <= 1.07: 0.169 (n = 353, err = 0.3)
| | | | | | | | [424] RAINFALL60 > 1.07: 0.203 (n = 201, err = 0.5)
| | | | [425] MEDHHINC > 77947.25845
| | | | | [426] PCUNEMPLOYED <= 2.4
| | | | | | [427] PCBORNUSA <= 95.5
| | | | | | | [428] MEDIANAGE <= 38.9
| | | | | | | | [429] SQMLAWN <= 898253.00126: 0.159 (n = 1189, err = 1.1)
| | | | | | | | [430] SQMLAWN > 898253.00126: 0.210 (n = 2510, err = 5.5)
| | | | | | | [431] MEDIANAGE > 38.9
| | | | | | | | [432] MEDCONSTYR <= 1986: 0.260 (n = 1288, err = 2.5)
| | | | | | | | [433] MEDCONSTYR > 1986: 0.217 (n = 2752, err = 3.4)
| | | | | | [434] PCBORNUSA > 95.5
| | | | | | | [435] FORECLOSED <= 17
| | | | | | | | [436] MEDCONSTYR <= 1984: 0.377 (n = 151, err = 0.4)
| | | | | | | | [437] MEDCONSTYR > 1984: 0.230 (n = 364, err = 1.2)
| | | | | | | [438] FORECLOSED > 17
| | | | | | | | [439] DATE <= 14322: 0.276 (n = 17, err = 0.0)
| | | | | | | | [440] DATE > 14322: 0.177 (n = 102, err = 0.2)
| | | | | [441] PCUNEMPLOYED > 2.4
| | | | | | [442] SQMLAWN <= 1925796.41343
| | | | | | | [443] SQMLAWN <= 907426.56864
| | | | | | | | [444] MEDIANAGE <= 28.7: 0.196 (n = 500, err = 0.5)
| | | | | | | | [445] MEDIANAGE > 28.7: 0.170 (n = 1723, err = 1.2)
| | | | | | | [446] SQMLAWN > 907426.56864
| | | | | | | | [447] MEDCONSTYR <= 1998: 0.218 (n = 3160, err = 6.2)
| | | | | | | | [448] MEDCONSTYR > 1998: 0.186 (n = 3591, err = 4.0)
| | | | | | [449] SQMLAWN > 1925796.41343
| | | | | | | [450] FORECLOSED <= 15
| | | | | | | | [451] MEDCONSTYR <= 1962: 0.396 (n = 68, err = 0.1)
| | | | | | | | [452] MEDCONSTYR > 1962: 0.236 (n = 3789, err = 10.2)
| | | | | | | [453] FORECLOSED > 15
| | | | | | | | [454] GEOINDEX <= 742: 0.215 (n = 2415, err = 4.0)
| | | | | | | | [455] GEOINDEX > 742: 0.181 (n = 874, err = 0.9)
Number of inner nodes: 227
Number of terminal nodes: 228
> cor.test(prediction, party_data$NDVI)
Pearson's product-moment correlation
data: prediction and party_data$NDVI
t = 404.86, df = 280980, p-value < 2.2e-16
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
0.6046468 0.6093172
sample estimates:
cor
0.6069872
3.2 Tract Trend Analysis
In contrast to the atemporal panel analysis, trend analysis looked specifically at the NDVI trend in tracts over time.
The models have modestly good fits, albeit with a smaller number of variables that have high significance.
3.2.1 Linear Model
As shown in the graph below, the linear model fit was modestly strong (r2 = 0.305).
The three variables with high significance in the linear model are median household income (MEDHHINCOME), the overall NDVI level (INTERCEPT), and percent of residents born in the us (PCBORNUSA). Wealthier tracts tended to have increased vegetation over the analysis period, while tracts with high levels of vegetation tended to have decreased vegetation over the analysis period.
Linear model NDVI change prediction vs actual values
(analysis/tract-trend/linear-prediction.png)
lm(formula = SLOPE ~ ., data = lin_data)
Residuals:
Min 1Q Median 3Q Max
-6.0561 -0.3647 -0.0108 0.4180 2.7387
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -0.2733830 0.0597368 -4.576 5.41e-06 ***
MEDHHINC 0.3109458 0.0500174 6.217 7.83e-10 ***
INTERCEPT -0.3107912 0.0334308 -9.297 < 2e-16 ***
PCBORNUSA 0.1630100 0.0438277 3.719 0.000212 ***
GEOINDEX 0.0006132 0.0001170 5.239 2.02e-07 ***
SQMLAND 0.0299508 0.0322587 0.928 0.353426
MEDIANAGE 0.1264691 0.0619903 2.040 0.041634 *
MEANHHSIZE 0.0761501 0.0587062 1.297 0.194923
PCOWNEROCC -0.1499318 0.0660742 -2.269 0.023501 *
PCTURNOVER -0.0140573 0.0418814 -0.336 0.737218
PCUNEMPLOYED -0.0513234 0.0361559 -1.420 0.156106
MEDCONSTYR 0.0189103 0.0402697 0.470 0.638764
SQMLAWN 0.1165299 0.0716808 1.626 0.104376
FDAYSQM -0.1214984 0.0668359 -1.818 0.069425 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.8396 on 881 degrees of freedom
(4 observations deleted due to missingness)
Multiple R-squared: 0.3051, Adjusted R-squared: 0.2949
F-statistic: 29.76 on 13 and 881 DF, p-value: < 2.2e-16
3.2.2 Generalized Additive Model
As shown in the graph below, the generalized additive model had a slightly better fit than the linear model (r2 = 0.391).
Two variables were found to have high significance: median household income (MEDHHINCOME) and median construction year of homes in the tract (MEDCONSTYR), although only median household income has a clearly positive relationship in the smoothing function.
Generalized additive model NDVI change prediction vs actual values
(analysis/tract-trend/gam-prediction.png)
Family: gaussian
Link function: identity
Formula:
SLOPE ~ s(SQMLAND) + s(MEDIANAGE) + s(MEDHHINC) + s(MEANHHSIZE) +
s(PCOWNEROCC) + s(PCTURNOVER) + s(PCBORNUSA) + s(PCUNEMPLOYED) +
s(SQMLAWN) + s(MEDCONSTYR) + s(FDAYSQM) + s(GEOINDEX)
Parametric coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -5.526e-07 2.053e-07 -2.691 0.00725 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Approximate significance of smooth terms:
edf Ref.df F p-value
s(SQMLAND) 1.000 1.000 0.466 0.495
s(MEDIANAGE) 1.483 1.834 2.942 0.058 .
s(MEDHHINC) 1.558 1.954 11.591 1.48e-05 ***
s(MEANHHSIZE) 4.720 5.829 1.224 0.291
s(PCOWNEROCC) 2.635 3.330 1.927 0.116
s(PCTURNOVER) 1.000 1.001 1.356 0.245
s(PCBORNUSA) 1.639 2.052 1.566 0.208
s(PCUNEMPLOYED) 1.000 1.001 0.779 0.378
s(SQMLAWN) 1.025 1.049 0.050 0.835
s(MEDCONSTYR) 8.832 8.988 17.796 < 2e-16 ***
s(FDAYSQM) 5.179 6.346 1.357 0.225
s(GEOINDEX) 7.422 8.380 7.350 9.70e-10 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
R-sq.(adj) = 0.391 Deviance explained = 41.6%
GCV = 3.9422e-11 Scale est. = 3.7727e-11 n = 895
Generalized additive model smoothing functions
(analysis/tract-trend/panel-gam-smooth.png)
3.2.3 Random Forest Model
The random forest model lists overall NDVI level (INTERCEPT) as the most important predictor variable, followed by median household income (MEDHHINC), percent born in the US (PCBORNUSA) and median age (MEDIANAGE).
Random forest model NDVI change prediction vs actual values
(analysis/tract-trend/rf-prediction.png)
> importance(rf_model)
IncNodePurity
INTERCEPT 6.967133e-09
GEOINDEX 6.079780e-09
MEDHHINC 4.986954e-09
PCBORNUSA 3.652985e-09
MEDIANAGE 3.148562e-09
MEANHHSIZE 3.024757e-09
SQMLAND 2.345369e-09
FDAYSQM 2.336801e-09
SQMLAWN 2.192065e-09
PCUNEMPLOYED 2.022879e-09
PCOWNEROCC 1.988081e-09
PCTURNOVER 1.474877e-09
MEDCONSTYR 1.245822e-08
3.2.4 Conditional Inference Tree (Party)
Conditional inference tree NDVI change prediction vs actual values
(analysis/tract-trend/party-prediction.png)
Model formula:
~SLOPE + INTERCEPT + (SQMLAND + MEDIANAGE + MEDHHINC + MEANHHSIZE +
PCOWNEROCC + PCTURNOVER + PCBORNUSA + PCUNEMPLOYED + MEDCONSTYR +
SQMLAWN + GEOINDEX + FDAYSQM)
Fitted party:
[1] root
| [2] PCUNEMPLOYED <= 6.2
| | [3] MEDCONSTYR <= 1995
| | | [4] MEDHHINC <= 76231
| | | | [5] MEDCONSTYR <= 1965
| | | | | [6] MEDHHINC <= 39786: *
| | | | | [7] MEDHHINC > 39786
| | | | | | [8] GEOINDEX <= 466: *
| | | | | | [9] GEOINDEX > 466: *
| | | | [10] MEDCONSTYR > 1965
| | | | | [11] MEDIANAGE <= 61.8
| | | | | | [12] MEDCONSTYR <= 1986
| | | | | | | [13] PCBORNUSA <= 88.9: *
| | | | | | | [14] PCBORNUSA > 88.9: *
| | | | | | [15] MEDCONSTYR > 1986
| | | | | | | [16] MEANHHSIZE <= 2.93: *
| | | | | | | [17] MEANHHSIZE > 2.93: *
| | | | | [18] MEDIANAGE > 61.8: *
| | | [19] MEDHHINC > 76231
| | | | [20] MEDCONSTYR <= 1993
| | | | | [21] MEDCONSTYR <= 1972: *
| | | | | [22] MEDCONSTYR > 1972
| | | | | | [23] MEDIANAGE <= 42.6: *
| | | | | | [24] MEDIANAGE > 42.6: *
| | | | [25] MEDCONSTYR > 1993
| | | | | [26] MEANHHSIZE <= 2.95: *
| | | | | [27] MEANHHSIZE > 2.95: *
| | [28] MEDCONSTYR > 1995
| | | [29] MEDHHINC <= 73512
| | | | [30] MEDIANAGE <= 58.2: *
| | | | [31] MEDIANAGE > 58.2: *
| | | [32] MEDHHINC > 73512
| | | | [33] MEDCONSTYR <= 2005
| | | | | [34] MEDCONSTYR <= 2001: *
| | | | | [35] MEDCONSTYR > 2001: *
| | | | [36] MEDCONSTYR > 2005: *
| [37] PCUNEMPLOYED > 6.2
| | [38] PCBORNUSA <= 82.5
| | | [39] MEDIANAGE <= 27.9
| | | | [40] GEOINDEX <= 374
| | | | | [41] SQMLAND <= 244363.56711
| | | | | | [42] MEDCONSTYR <= 1987: *
| | | | | | [43] MEDCONSTYR > 1987: *
| | | | | [44] SQMLAND > 244363.56711: *
| | | | [45] GEOINDEX > 374: *
| | | [46] MEDIANAGE > 27.9: *
| | [47] PCBORNUSA > 82.5
| | | [48] MEDCONSTYR <= 1992: *
| | | [49] MEDCONSTYR > 1992
| | | | [50] GEOINDEX <= 111: *
| | | | [51] GEOINDEX > 111: *
Number of inner nodes: 25
Number of terminal nodes: 26
Appendix: Variable Correlation
The list below shows the correlation between examined variables, sorted in descending absolute value order. When multiple variables were highly correlated to each other, only one was chosen to avoid confusing the models. The choices were based on presumed dependent relationship; for example, MEDHHINC was chosen over MEDHOMEPRICE since income limits the maximum home price a buyer can afford rather than the other way around. These choices are contestible.
Excluded variables (r >= 0.7):
- MEDHOMEAGE
- MEDHOMEPRICE
- MEDHOMEVALUE
- MEDSQMHOME
- PCCOLLEGE
- PCLIVEALONE
PCVACANT (222300) and HOUSINGUNITS (335481) excluded because of large amount of missing data (not computed from ACS?)
x y r
1 MEDCONSTYR MEDHOMEAGE 0.98
2 MEDHOMEVALUE MEDHOMEPRICE 0.839
3 MEDSQMHOME MEDHOMEPRICE 0.831
4 MEDHHINC MEDSQMHOME 0.814
5 MEDHHINC MEDHOMEVALUE 0.767
6 MEDHHINC MEDHOMEPRICE 0.752
7 MEDHHINC PCCOLLEGE 0.75
8 PCCOLLEGE MEDHOMEVALUE 0.749
9 MEANHHSIZE PCLIVEALONE -0.741
10 PCCOLLEGE MEDHOMEPRICE 0.7
11 MEDHOMEVALUE MEDSQMHOME 0.694
12 MEDIANAGE MEANHHSIZE -0.629
13 PCCOLLEGE MEDSQMHOME 0.592
14 PCOWNEROCC PCTURNOVER 0.561
15 PCCOLLEGE PCBORNUSA 0.532
16 MEDHHINC PCBORNUSA 0.517
17 SQMLAWN MEDHOMEPRICE 0.514
18 MEDSQMHOME MEDCONSTYR 0.501
19 MEDSQMHOME MEDHOMEAGE 0.492
20 MEDHHINC MEDCONSTYR 0.486
21 MEDHHINC MEDHOMEAGE 0.486
22 SQMLAND SQMWATER 0.482
23 PCBORNUSA MEDCONSTYR 0.48
24 PCBORNUSA MEDHOMEAGE 0.478
25 MEDIANAGE PCBORNUSA 0.468
26 MEDHHINC PCLIVEALONE -0.466
27 SQMLAWN MEDSQMHOME 0.462
28 MEANHHSIZE PCBORNUSA -0.458
29 PCBORNUSA MEDSQMHOME 0.458
30 PCLIVEALONE MEDSQMHOME -0.452
31 MEDHOMEVALUE SQMLAWN 0.443
32 MEDIANAGE PCVACANT 0.417
33 SQMLAND SQMLAWN 0.414
34 PCBORNUSA PCUNEMPLOYED -0.414
35 MEANHHSIZE PCCOLLEGE -0.407
36 PCBORNUSA MEDHOMEPRICE 0.4
37 MEDHHINC SQMLAWN 0.391
38 PCBORNUSA MEDHOMEVALUE 0.381
39 MEDHHINC PCUNEMPLOYED -0.378
40 PCUNEMPLOYED MEDHOMEAGE -0.366
41 MEDIANAGE PCCOLLEGE 0.363
42 MEDCONSTYR MEDHOMEPRICE 0.358
43 PCLIVEALONE MEDHOMEAGE -0.355
44 MEDHOMEAGE MEDHOMEPRICE 0.349
45 MEDIANAGE PCUNEMPLOYED -0.347
46 MEANHHSIZE PCVACANT -0.336
47 PCLIVEALONE MEDCONSTYR -0.332
48 PCCOLLEGE PCUNEMPLOYED -0.327
49 PCUNEMPLOYED MEDSQMHOME -0.325
50 MEDHOMEVALUE MEDCONSTYR 0.322
51 MEDIANAGE MEDHOMEPRICE 0.321
52 PCCOLLEGE MEDCONSTYR 0.321
53 PCUNEMPLOYED MEDHOMEPRICE -0.32
54 SQMWATER SQMLAWN 0.316
55 MEDIANAGE MEDHOMEVALUE 0.314
56 PCTURNOVER PCUNEMPLOYED 0.307
57 MEDHOMEVALUE MEDHOMEAGE 0.304
58 PCCOLLEGE MEDHOMEAGE 0.294
59 PCLIVEALONE PCVACANT 0.276
60 PCUNEMPLOYED MEDCONSTYR -0.275
61 MEDIANAGE PCOWNEROCC 0.273
62 PCUNEMPLOYED MEDHOMEVALUE -0.273
63 MEDIANAGE PCLIVEALONE 0.272
64 NDVI MEDCONSTYR -0.269
65 MEDIANAGE SQMLAWN 0.269
66 NDVI MEDHOMEAGE -0.262
67 PCCOLLEGE SQMLAWN 0.261
68 HOUSINGUNITS MEDHOMEAGE 0.26
69 MEANHHSIZE PCUNEMPLOYED 0.255
70 MEDIANAGE PCTURNOVER 0.246
71 HOUSINGUNITS MEDCONSTYR 0.244
72 NDVI PCCOLLEGE 0.243
73 PCTURNOVER MEDHOMEVALUE 0.235
74 MEDIANAGE MEDSQMHOME 0.229
75 PCBORNUSA SQMLAWN 0.229
76 PCOWNEROCC PCCOLLEGE 0.224
77 SQMWATER MEDHOMEPRICE 0.22
78 NDVI PCVACANT -0.217
79 SQMLAWN MEDCONSTYR 0.215
80 MEDIANAGE MEDHHINC 0.214
81 PCOWNEROCC SQMLAWN 0.208
82 PCLIVEALONE SQMLAWN -0.204
83 SQMLAWN MEDHOMEAGE 0.204
84 NDVI HOUSINGUNITS -0.186
85 PCLIVEALONE MEDHOMEPRICE -0.182
86 NDVI MEDHOMEVALUE 0.18
87 MEANHHSIZE MEDHOMEVALUE -0.178
88 PCTURNOVER MEDHOMEAGE -0.177
89 NDVI MEDHOMEPRICE 0.175
90 MEANHHSIZE MEDHOMEPRICE -0.175
91 PCOWNEROCC PCUNEMPLOYED 0.171
92 PCOWNEROCC MEDHOMEVALUE 0.167
93 PCOWNEROCC PCBORNUSA 0.161
94 PCUNEMPLOYED SQMLAWN -0.161
95 PCOWNEROCC MEDSQMHOME 0.154
96 MEDHHINC PCOWNEROCC 0.153
97 PCOWNEROCC MEDCONSTYR 0.152
98 PCVACANT MEDCONSTYR 0.151
99 HOUSINGUNITS MEDSQMHOME 2.148
100 PCVACANT HOUSINGUNITS 0.146
101 PCVACANT MEDHOMEAGE 0.142
102 HOUSINGUNITS PCBORNUSA 0.137
103 HOUSINGUNITS MEDHOMEVALUE 0.135
104 MEDHHINC PCVACANT -0.134
105 PCOWNEROCC MEDHOMEPRICE 0.134
106 PCLIVEALONE MEDHOMEVALUE -0.132
107 NDVI MEDHHINC 0.13
108 PCCOLLEGE PCTURNOVER 0.13
109 SQMWATER MEDSQMHOME 0.127
110 PCVACANT PCUNEMPLOYED -0.126
111 MEDIANAGE MEDCONSTYR 0.124
112 SQMWATER MEDHOMEVALUE 0.123
113 PCTURNOVER SQMLAWN 0.123
114 PCVACANT MEDHOMEVALUE 0.12
115 HOUSINGUNITS SQMLAWN 0.12
116 SQMWATER PCVACANT 0.119
117 MEDHHINC HOUSINGUNITS 0.114
118 PCLIVEALONE PCUNEMPLOYED 0.114
119 MEDHHINC PCTURNOVER 0.113
120 NDVI PCLIVEALONE 0.107
121 SQMLAND PCVACANT 0.107
122 SQMLAND MEDHOMEPRICE 0.107
123 MEDIANAGE MEDHOMEAGE 0.105
124 HOUSINGUNITS MEDHOMEPRICE 0.104
125 HOUSINGUNITS PCTURNOVER -0.097
126 NDVI MEDSQMHOME 0.096
127 PCCOLLEGE HOUSINGUNITS 0.096
128 PCVACANT SQMLAWN 0.092
129 PCOWNEROCC HOUSINGUNITS -0.088
130 MEANHHSIZE PCOWNEROCC -0.086
131 PCVACANT MEDHOMEPRICE 0.084
132 PCTURNOVER MEDSQMHOME 0.081
133 NDVI MEANHHSIZE -0.08
134 SQMWATER MEDIANAGE 0.08
135 SQMWATER MEDHHINC 0.079
136 SQMWATER PCCOLLEGE 0.071
137 SQMLAND MEDSQMHOME 0.069
138 SQMLAND MEDCONSTYR 0.066
139 PCTURNOVER MEDHOMEPRICE 0.066
140 NDVI SQMLAWN 0.065
141 SQMLAND MEDHOMEAGE 0.065
142 MEANHHSIZE MEDSQMHOME 0.065
143 SQMLAND MEDIANAGE 0.064
144 PCVACANT PCOWNEROCC -0.058
145 HOUSINGUNITS PCUNEMPLOYED -0.058
146 PCVACANT PCBORNUSA 0.057
147 MEDIANAGE HOUSINGUNITS 0.055
148 SQMLAND PCLIVEALONE -0.053
149 SQMLAND MEDHOMEVALUE 0.051
150 MEANHHSIZE PCTURNOVER 0.051
151 NDVI PCBORNUSA 0.05
152 PCLIVEALONE PCTURNOVER 0.045
153 SQMWATER MEDCONSTYR 0.043
154 SQMWATER MEDHOMEAGE 0.042
155 NDVI MEDIANAGE -0.041
156 NDVI PMINPET 0.041
157 SQMWATER MEANHHSIZE -0.041
158 PCLIVEALONE PCCOLLEGE 0.039
159 PCTURNOVER PMINPET -0.039
160 PCVACANT MEDSQMHOME -0.038
161 HOUSINGUNITS PMINPET -0.038
162 MEDHHINC MEANHHSIZE 0.036
163 MEANHHSIZE SQMLAWN -0.034
164 PCLIVEALONE HOUSINGUNITS -0.034
165 SQMWATER PCLIVEALONE -0.033
166 SQMWATER PCUNEMPLOYED -0.033
167 PCTURNOVER MEDCONSTYR -0.032
168 NDVI SQMLAND -0.03
169 SQMLAND PCUNEMPLOYED -0.03
170 SQMWATER PCOWNEROCC 0.03
171 NDVI PCTURNOVER 0.027
172 SQMWATER PCBORNUSA 0.027
173 SQMLAND PCCOLLEGE -0.025
174 SQMWATER HOUSINGUNITS 0.025
175 PCVACANT PCCOLLEGE -0.025
176 PCOWNEROCC MEDHOMEAGE 0.023
177 PCLIVEALONE PCOWNEROCC 0.022
178 PCTURNOVER PCBORNUSA 0.021
179 MEANHHSIZE HOUSINGUNITS -0.02
180 MEANHHSIZE MEDCONSTYR -0.02
181 PCVACANT PCTURNOVER -0.02
182 PCOWNEROCC PMINPET -0.02
183 PCUNEMPLOYED PMINPET -0.019
184 SQMLAND PCBORNUSA 0.018
185 MEANHHSIZE MEDHOMEAGE -0.018
186 NDVI PCUNEMPLOYED -0.017
187 NDVI PCOWNEROCC -0.016
188 SQMLAND HOUSINGUNITS 0.015
189 SQMLAND MEANHHSIZE -0.014
190 SQMLAND PCOWNEROCC 0.014
191 SQMLAND PCTURNOVER 0.013
192 MEDHOMEVALUE PMINPET -0.013
193 PCVACANT PMINPET -0.012
194 SQMLAND MEDHHINC 0.007
195 SQMWATER PCTURNOVER 0.007
196 PCLIVEALONE PMINPET -0.007
197 PCCOLLEGE PMINPET -0.005
198 MEDHOMEAGE PMINPET 0.005
199 MEANHHSIZE PMINPET -0.004
200 MEDIANAGE PMINPET -0.003
201 MEDCONSTYR PMINPET -0.003
202 NDVI SQMWATER 0.002
203 MEDHHINC PMINPET -0.002
204 PCLIVEALONE PCBORNUSA -0.002
205 SQMLAWN PMINPET -0.002
206 PCBORNUSA PMINPET 0.001
207 MEDSQMHOME PMINPET -0.001
208 MEDHOMEPRICE PMINPET -0.001
209 SQMLAND PMINPET 0
210 SQMWATER PMINPET 0