DHS Wealth Index

Basic introduction to processing DHS data

Loading DHS Data

Note: This data is mocked data. Request official data from their available datasets through their website.

import geopandas as gpd
import pandas as pd

from geowrangler import dhs
dhs_household_data_path = "../data/ph.DTA"
dhs_gps_coordinates = "../data/ph_gps.shp"
dhs_df = dhs.load_dhs_file(dhs_household_data_path)
CPU times: user 11.2 ms, sys: 583 ยตs, total: 11.8 ms
Wall time: 10.6 ms
dhs_df.head()
country code and phase cluster number source of drinking water type of toilet facility has electricity has radio has television has refrigerator has motorcycle/scooter has car/truck main floor material number of rooms used for sleeping has mobile telephone wealth index factor score combined (5 decimals)
0 0 PH7 725 80 61 0 1 0 1 1 1 74 20 0 -168581
1 1 PH7 1009 19 92 1 1 1 1 1 0 51 15 1 127550
2 2 PH7 1072 91 58 1 1 0 1 1 1 26 5 0 32616
3 3 PH7 242 39 93 0 1 0 0 0 1 76 11 0 80338
4 4 PH7 102 11 27 0 0 1 1 1 0 56 15 0 -178758

Renaming columns to match

DHS files do not have uniform column names. To make analysis easier, we rename commonnly ones. Feel free to extend the config to your usecase.

ph_config = dhs.load_column_config("ph")
ph_config
{'cluster number': 'DHSCLUST',
 'wealth index factor score combined (5 decimals)': 'Wealth Index',
 'country code and phase': 'country code and phase',
 'number of rooms used for sleeping': 'rooms',
 'has electricity': 'electric',
 'has mobile telephone': 'mobile telephone',
 'has radio': 'radio',
 'has television': 'television',
 'has car/truck': 'car/truck',
 'has refrigerator': 'refrigerator',
 'has motorcycle/scooter': 'motorcycle',
 'main floor material': 'floor',
 'type of toilet facility': 'toilet',
 'source of drinking water': 'drinking water'}
renamed_dhs_df = dhs_df.rename(columns=ph_config)
renamed_dhs_df.head()
country code and phase DHSCLUST drinking water toilet electric radio television refrigerator motorcycle car/truck floor rooms mobile telephone Wealth Index
0 0 PH7 725 80 61 0 1 0 1 1 1 74 20 0 -168581
1 1 PH7 1009 19 92 1 1 1 1 1 0 51 15 1 127550
2 2 PH7 1072 91 58 1 1 0 1 1 1 26 5 0 32616
3 3 PH7 242 39 93 0 1 0 0 0 1 76 11 0 80338
4 4 PH7 102 11 27 0 0 1 1 1 0 56 15 0 -178758

Cluster Summaries

wealth_col_name = "Wealth Index"
cluster_col_name = "DHSCLUST"
summarized = (
    renamed_dhs_df[[wealth_col_name, cluster_col_name]].groupby(cluster_col_name).mean()
)
summarized = summarized.reset_index()
summarized.head()
DHSCLUST Wealth Index
0 3 -232188.0
1 4 228860.0
2 5 157620.0
3 6 40308.5
4 7 -37595.5 
len(summarized)
679
summarized.DHSCLUST.isna().sum()
0
summarized.DHSCLUST.isna().sum()
0
dhs_shp = gpd.read_file(dhs_gps_coordinates)
dhs_shp.head()
DHSID DHSCC DHSYEAR DHSCLUST LATNUM LONGNUM geometry
0 PH20XX725 PH 0 725 0.409441 0.220510 POINT (0.22051 0.40944)
1 PH20XX1009 PH 0 1009 0.333693 0.332499 POINT (0.33250 0.33369)
2 PH20XX1072 PH 0 1072 0.378053 0.089852 POINT (0.08985 0.37805)
3 PH20XX242 PH 0 242 0.306277 0.431677 POINT (0.43168 0.30628)
4 PH20XX102 PH 0 102 0.535456 0.716025 POINT (0.71602 0.53546) 
survey_geo = pd.merge(summarized, dhs_shp, on="DHSCLUST")
survey_geo
DHSCLUST Wealth Index DHSID DHSCC DHSYEAR LATNUM LONGNUM geometry
0 3 -232188.000000 PH20XX003 PH 0 0.609945 0.350830 POINT (0.35083 0.60995)
1 4 228860.000000 PH20XX004 PH 0 0.363843 0.281563 POINT (0.28156 0.36384)
2 5 157620.000000 PH20XX005 PH 0 0.715438 0.145014 POINT (0.14501 0.71544)
3 6 40308.500000 PH20XX006 PH 0 0.758501 0.628373 POINT (0.62837 0.75850)
4 6 40308.500000 PH20XX006 PH 0 0.669415 0.479379 POINT (0.47938 0.66942)
... ... ... ... ... ... ... ... ...
995 1247 90273.333333 PH20XX1247 PH 0 0.214682 0.419448 POINT (0.41945 0.21468)
996 1248 212729.500000 PH20XX1248 PH 0 0.189401 0.152806 POINT (0.15281 0.18940)
997 1248 212729.500000 PH20XX1248 PH 0 0.563460 0.023900 POINT (0.02390 0.56346)
998 1250 101508.500000 PH20XX1250 PH 0 0.166465 0.617584 POINT (0.61758 0.16646)
999 1250 101508.500000 PH20XX1250 PH 0 0.252438 0.668243 POINT (0.66824 0.25244)

1000 rows ร— 8 columns

Recalculating wealth index for a single country

features = [
    "rooms",
    "electric",
    "mobile telephone",
    "radio",
    "television",
    "car/truck",
    "refrigerator",
    "motorcycle",
    "floor",
    "toilet",
    "drinking water",
]
# apply a threshold
dhs.apply_threshold(renamed_dhs_df, columns=features, config={"rooms": [0, 25]})
country code and phase DHSCLUST drinking water toilet electric radio television refrigerator motorcycle car/truck floor rooms mobile telephone Wealth Index
0 0 PH7 725 80 61 0 1 0 1 1 1 74 20 0 -168581
1 1 PH7 1009 19 92 1 1 1 1 1 0 51 15 1 127550
2 2 PH7 1072 91 58 1 1 0 1 1 1 26 5 0 32616
3 3 PH7 242 39 93 0 1 0 0 0 1 76 11 0 80338
4 4 PH7 102 11 27 0 0 1 1 1 0 56 15 0 -178758
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
995 995 PH7 834 63 83 0 1 1 1 1 1 28 10 0 241846
996 996 PH7 1170 74 78 1 0 1 1 1 0 34 17 1 106424
997 997 PH7 762 81 14 1 1 0 1 1 0 83 12 1 246898
998 998 PH7 63 40 21 0 1 1 1 1 1 12 2 1 -212590
999 999 PH7 748 56 38 1 1 1 0 1 1 69 7 1 -10774

1000 rows ร— 15 columns

renamed_dhs_df["Recomputed Wealth Index"] = dhs.assign_wealth_index(
    renamed_dhs_df[features], features
)
renamed_dhs_df.head()
country code and phase DHSCLUST drinking water toilet electric radio television refrigerator motorcycle car/truck floor rooms mobile telephone Wealth Index Recomputed Wealth Index
0 0 PH7 725 80 61 0 1 0 1 1 1 74 20 0 -168581 -63.515731
1 1 PH7 1009 19 92 1 1 1 1 1 0 51 15 1 127550 8.730482
2 2 PH7 1072 91 58 1 1 0 1 1 1 26 5 0 32616 -60.121255
3 3 PH7 242 39 93 0 1 0 0 0 1 76 11 0 80338 -14.959946
4 4 PH7 102 11 27 0 0 1 1 1 0 56 15 0 -178758 -14.060453

Recalculating wealth index for multiple countries

dhs_ph_path = "../data/ph.DTA"
dhs_kh_path = "../data/kh.DTA"
dhs_mm_path = "../data/mm.DTA"
dhs_tl_path = "../data/tl.DTA"
ph_config = dhs.load_column_config("ph")
kh_config = dhs.load_column_config("kh")
mm_config = dhs.load_column_config("mm")
tl_config = dhs.load_column_config("tl")
dhs_ph_df = dhs.load_dhs_file(dhs_ph_path).rename(columns=ph_config)
dhs_kh_df = dhs.load_dhs_file(dhs_kh_path).rename(columns=kh_config)
dhs_mm_df = dhs.load_dhs_file(dhs_mm_path).rename(columns=mm_config)
dhs_tl_df = dhs.load_dhs_file(dhs_tl_path).rename(columns=tl_config)
cols = list(ph_config.values()) + ["country code and phase"]
merged_df = pd.concat(
    [
        dhs_ph_df[cols],
        dhs_kh_df[cols],
        dhs_mm_df[cols],
        dhs_tl_df[cols],
    ]
)
merged_df = merged_df.fillna(0)
merged_df["Recomputed Wealth Index"] = dhs.assign_wealth_index(merged_df[features])
merged_df.head()
DHSCLUST Wealth Index country code and phase rooms electric mobile telephone radio television car/truck refrigerator motorcycle floor toilet drinking water country code and phase Recomputed Wealth Index
0 725 -168581 PH7 20 0 0 1 0 1 1 1 74 61 80.0 PH7 122.888238
1 1009 127550 PH7 15 1 1 1 1 0 1 1 51 92 19.0 PH7 78.430888
2 1072 32616 PH7 5 1 0 1 0 1 1 1 26 58 91.0 PH7 108.113989
3 242 80338 PH7 11 0 0 1 0 1 0 0 76 93 39.0 PH7 105.464687
4 102 -178758 PH7 15 0 0 0 1 0 1 1 56 27 11.0 PH7 45.993876 
import matplotlib.pyplot as plt  # noqa
merged_df.hist("Recomputed Wealth Index")
array([[<Axes: title={'center': 'Recomputed Wealth Index'}>]],
      dtype=object)

merged_df["Recomputed Wealth Index Not PCA"] = dhs.assign_wealth_index(
    merged_df[features], use_pca=False
)
merged_df.head()
DHSCLUST Wealth Index country code and phase rooms electric mobile telephone radio television car/truck refrigerator motorcycle floor toilet drinking water country code and phase Recomputed Wealth Index Recomputed Wealth Index Not PCA
0 725 -168581 PH7 20 0 0 1 0 1 1 1 74 61 80.0 PH7 122.888238 -10730.578127
1 1009 127550 PH7 15 1 1 1 1 0 1 1 51 92 19.0 PH7 78.430888 -57116.290440
2 1072 32616 PH7 5 1 0 1 0 1 1 1 26 58 91.0 PH7 108.113989 30004.584457
3 242 80338 PH7 11 0 0 1 0 1 0 0 76 93 39.0 PH7 105.464687 -48966.528484
4 102 -178758 PH7 15 0 0 0 1 0 1 1 56 27 11.0 PH7 45.993876 -39722.455483 
merged_df.hist("Recomputed Wealth Index Not PCA")
array([[<Axes: title={'center': 'Recomputed Wealth Index Not PCA'}>]],
      dtype=object)