Spaces:
Runtime error
Runtime error
File size: 48,085 Bytes
5ed3053 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 | ## MOPAC Spatial Propensity Matching
## Daniel Hammocks - 2024-08-14
## GH: dhammo2
## This code takes the JSON extracted from the Metropolitan Police R HTML Ouput
## and creates a geoJSON/Shapefile of control and intervention groups including
## all data displayed on the graph.
###############################################################################
################################# APPLICATION #################################
###############################################################################
#%% Versioning
# v1.0.0 - Original Dashboard
# v2.0.0 - Introduced Download Option
# v3.0.0 - Added Multi-Page Functionality
# v4.0.0 - Introduced Quality Assurance
# - Altered Calculation for PSM to search control only.
# v5.0.0 - Ensure that Area Matching Does Not Return Self
# v6.0.0 - Added Quality Assurance
# v7.0.0 - Complete Quality Assurance Information
#%% NOTES
#DISABLE LOGGING WHEN DONE
#Put Error Notices for QA at top.
#%% Required Libraries
import dash
import dash_bootstrap_components as dbc
from dash import dcc, html, Input, Output, State
import pandas as pd
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import NearestNeighbors
from sklearn.preprocessing import StandardScaler
import io
import base64
from datetime import datetime
from scipy import stats
import plotly.graph_objects as go
from scipy.stats import gaussian_kde
import logging
import os
#%% App Initialisation
# Initialize the Dash app
app = dash.Dash(__name__, suppress_callback_exceptions=True, external_stylesheets=[dbc.themes.BOOTSTRAP])
app.title = "MOPAC | DS"
logging.basicConfig(level=logging.DEBUG)
#%% Navigation Layout
navbar = dbc.NavbarSimple(
children=[
dbc.NavItem(dbc.NavLink("Overview", href="/")),
dbc.NavItem(dbc.NavLink("Area Similarity & Matching", href="/match")),
dbc.NavItem(dbc.NavLink("Quality Assurance", href="/quality")),
],
brand="MOPAC | DS - Area Similarity and Matching Application",
brand_href="#",
color="primary",
dark=True,
fluid=True
)
#%% Page Layout
overview_layout = html.Div(
children=[html.Br(),
html.H1(children="Overview"),
html.P("This application enables users to identify similar "
"geographic areas based on a set of covariates "
"(characteristics) related to each area. Two methods have "
"been implemented enabling the user to both identify "
"similar areas based on a set of characteristics but to "
"also identify similar areas once a treatment has already "
"been assigned. The two methods are detailed below."),
html.Br(),
html.Hr(),
html.H3('Data Formatting Requirements'),
html.Div([
html.Div([
html.H4('Example Data Structure'),
html.Br(),
html.Table([
html.Thead(
html.Tr([
html.Th('Area_ID', style={'border': '1px solid black'}),
html.Th('Population', style={'border': '1px solid black'}),
html.Th('Income', style={'border': '1px solid black'}),
html.Th('Education', style={'border': '1px solid black'}),
html.Th('Treatment (optional)', style={'border': '1px solid black'})
], style={'background-color': '#f2f2f2', 'font-weight': 'bold', 'text-align': 'center'})
),
html.Tbody([
html.Tr([
html.Td('Area_001', style={'border': '1px solid black'}),
html.Td('10000', style={'border': '1px solid black'}),
html.Td('50000', style={'border': '1px solid black'}),
html.Td('12', style={'border': '1px solid black'}),
html.Td('1', style={'border': '1px solid black'})
]),
html.Tr([
html.Td('Area_002', style={'border': '1px solid black'}),
html.Td('15000', style={'border': '1px solid black'}),
html.Td('60000', style={'border': '1px solid black'}),
html.Td('14', style={'border': '1px solid black'}),
html.Td('0', style={'border': '1px solid black'})
]),
html.Tr([
html.Td('Area_003', style={'border': '1px solid black'}),
html.Td('12000', style={'border': '1px solid black'}),
html.Td('55000', style={'border': '1px solid black'}),
html.Td('13', style={'border': '1px solid black'}),
html.Td('1', style={'border': '1px solid black'})
]),
html.Tr([
html.Td('...', style={'border': '1px solid black'}),
html.Td('...', style={'border': '1px solid black'}),
html.Td('...', style={'border': '1px solid black'}),
html.Td('...', style={'border': '1px solid black'}),
html.Td('...', style={'border': '1px solid black'})
])
], style={'text-align': 'center'}),
], style={
'width': '80%',
'border-collapse': 'collapse',
'margin-top': '10px',
'border': '1px solid black'
}),
html.Br(),
html.A(
html.Img(src='assets/DataOverview.png',
style={'width': '85%', 'height': 'auto'})
, href="assets/DataOverview.png", target="_blank", style={'font-size': '20px', 'text-decoration': 'none', 'color': 'blue'}),
html.P("Click to Expand.", style={'width': '85%', 'textAlign': 'center'})
], style={'width': '33%', 'display': 'inline-block', 'vertical-align': 'top'}),
html.Div([
html.H4('Required Data Format'),
html.P('To use the Area Similarity and Matching application, please provide your data in a CSV format with the following structure:'),
html.H5('1. Area ID/Name'),
html.P('A unique identifier or name for each area (e.g., "Area_ID" or "Area_Name"). This will be used to identify and match areas.'),
html.H5('2. Covariate Columns'),
html.P('One or more columns containing the demographic or other relevant variables (covariates) that will be used to compare areas. Examples include population, income, education level, etc.'),
html.H5('3. [Optional] Treatment Column'),
html.P('The treatment column is optional but can be included if you want to perform Propensity Score Matching (PSM). This should be a binary column indicating whether an area has received a treatment (e.g., "Treatment", where 1 = treated, 0 = not treated). If this column is provided, the application will use it to perform PSM. If not provided, the application will simply match areas based on similarity using covariates.'),
html.H4('Key Points'),
html.Ul([
html.Li('The Area ID/Name must be unique for each area.'),
html.Li('Covariates should be numeric and relevant to the analysis.'),
html.Li('The Treatment column is optional; if included, it allows for Propensity Score Matching, otherwise, the app can identify similar areas without considering treatment.')
])
], style={'width': '66%', 'display': 'inline-block', 'vertical-align': 'top'}),
]),
html.Hr(),
dbc.Row([
dbc.Col([
html.H3("Propensity Score Matching (PSM)"),
html.B("This method assumes the treatment areas are known.", style={'color':'red'}),
html.H5('What is the Goal?'),
html.P('The goal of PSM is to find areas that are most similar to each other based on certain covariates (characteristics like population, income, etc.) in order to balance the covariates between the treatment and control groups. By ensuring that the treatment and control groups are balanced in terms of covariates, PSM helps reduce confounding bias, leading to more accurate estimates of the treatment effect.'),
html.H5('Step 1: Collect Information'),
html.Ul([
html.Li('Start with a list of areas that received a "treatment" (e.g., a new program or policy) and potential control areas.'),
html.Li('Gather information (like demographics) about both treated and nearby areas.')
]),
html.H5('Step 2: Calculate a Score (Using Logistic Regression)'),
html.Ul([
html.Li('Use a statistical method called logistic regression to calculate a "score" for each area.'),
html.Li('This propensity score tells us how likely each area is to have received the treatment based on its characteristics.')
]),
html.H5('Step 3: Match Areas (Using Nearest Neighbor)'),
html.Ul([
html.Li('For each treated area, find a similar untreated area with the closest score.'),
html.Li('This is like finding the "nearest neighbor" in terms of similarity.')
]),
html.H5('Step 4: Identify Similar Areas'),
html.Ul([
html.Li('After matching, you have pairs of areas where each treated area is matched with a nearby area that is most similar based on the characteristics considered.')
]),
html.H5('Result'),
html.P('We now know which areas are most similar to each other, helping us understand the impact of the treatment by comparing them.')
], width = 6),
dbc.Col([
html.H3("Nearest Neighbour Matching (NNM)"),
html.H5('What is the Goal?'),
html.P('We want to find areas that are most similar to each other based on certain characteristics (like population, income, etc.), without focusing on any specific treatment.'),
html.H5('Step 1: Collect Information'),
html.Ul([
html.Li('Start by gathering information about the areas you are interested in, such as demographics, socio-economic data, or any other relevant characteristics.')
]),
html.H5('Step 2: Calculate Similarity'),
html.Ul([
html.Li('Use the characteristics (also called covariates) of each area to calculate a measure of similarity between areas.'),
html.Li('The most common approach is to calculate the "distance" between areas in the space defined by these covariates.')
]),
html.H5('Step 3: Find Nearest Neighbors'),
html.Ul([
html.Li('For each area, find the other area that has the smallest "distance" to it.'),
html.Li('This is known as finding the "nearest neighbor."'),
html.Li('The nearest neighbor is considered the most similar area based on the covariates.')
]),
html.H5('Step 4: Identify Similar Areas'),
html.Ul([
html.Li('After finding the nearest neighbors, you have a list of pairs of areas where each pair consists of the most similar areas based on the characteristics considered.')
]),
html.H5('Result'),
html.P('You now know which areas are most similar to each other, based purely on their characteristics, allowing for various analyses or comparisons.')
], width = 6)
])
]
)
match_layout = html.Div(
children=[
html.Br(),
html.H1("Area Similarity and Matching", className="mb-2"),
dbc.Row([
dbc.Col(html.P("Upload a CSV file with your geographic areas and demographic variables. "
"Select the analysis mode below and follow the steps to identify similar areas."),
className="mb-4")
]),
dbc.Row([
dbc.Col([
html.H4("Settings"),
dbc.Col(html.H5("Treatment Status"), className="mb-2"),
dbc.Col(html.P("Have the treatment areas already been selected?" ),
className="mb-4"),
dbc.RadioItems(
options=[
{"label": "Yes", "value": 1},
{"label": "No", "value": 0}
],
value=0, # Default value
id="treatment-toggle",
inline=True, # Display radio buttons inline
className="mb-4"
),
dbc.Col(html.H5("File Upload"), className="mb-2"),
dcc.Upload(
id='upload-data',
children=html.Div([
'Drag and Drop or ',
html.A('Select Files')
]),
style={
'width': '100%', 'height': '60px', 'lineHeight': '60px',
'borderWidth': '1px', 'borderStyle': 'dashed',
'borderRadius': '5px', 'textAlign': 'center', 'margin': '10px',
'margin-left' : '0px'
},
multiple=False
),
dbc.Progress(id="upload-progress",
value=0,
striped=True,
animated=True,
style={'transition': 'width 0.5s',
'backgroundColor': 'lightgray'},
className="mb-4"),
html.Div(id='output-data-upload'),
dbc.Col(html.H5("Column Selection"), className="mb-2"),
html.H6("Select Area ID/Name Column"),
dcc.Dropdown(id='area-id-column', placeholder="Select Area ID/Name Column"),
html.H6("Select Treatment Column (if applicable)"),
dcc.Dropdown(id='treatment-column', placeholder="Select Treatment Column"),
html.H6("Select Covariate Columns"),
dcc.Dropdown(id='covariate-columns', multi=True, placeholder="Select Covariate Columns"),
html.Br(),
html.Button("Run Analysis", id="run-analysis", n_clicks=0, className="btn btn-primary"),
html.Button("Clear Data", id="clear-button", className="btn btn-primary",
style={
'backgroundColor': 'red', # Red background color
'color': 'white', # White text color
'border': 'none', # Remove border
'cursor': 'pointer',
'margin-left': '20px'
}),
html.Br(),
html.Br(),
html.H4("Download Results"),
dcc.Store(id='raw-results'),
html.Div(html.Button("Download Results",
id = "download-button",
n_clicks = 0,
className="btn btn-secondary"),
style={'margin-top': '10px'}),
dcc.Download(id="download-data")
], width=3),
dbc.Col([
html.H4("Output"),
html.Div(id="analysis-output")
], width=9)
])
]
)
quality_layout = html.Div([
html.H1("Propensity Score Matching Evaluation Metrics"),
html.Br(),
dbc.Row([
dbc.Col([
html.Button("Run Quality Assurance Script", id="run-quality", n_clicks=0, className="btn btn-primary", style = {'margin-bottom': '20px',}),
], width = 3),
dbc.Col([
html.Div(id='output-quality-error', style={'color':'red'})
], width = 9)
]),
html.Br(),
dbc.Row([
html.H3("Distribution of Propensity Scores"),
dbc.Col([
html.H4("Plot"),
html.Div(id='propensity-distribution')
], width=8),
dbc.Col([
html.H4("Explanation"),
html.P(
"In this application, the Propensity Score Distribution visualises the likelihood that an area "
"or individual would receive the treatment based on observed covariates. The propensity score "
"is a value between 0 and 1 that indicates the probability of receiving treatment, given a set "
"of characteristics. "
"The plot displays the distribution of propensity scores "
"for both the treated and control groups. "
),
html.P("Key Considerations:"),
html.Ul([
html.Li(
"Overlap of Distributions: Ideally, the propensity score distributions for treated and control "
"groups should overlap significantly. This overlap indicates that the groups are comparable in terms "
"of covariates, which is crucial for reducing bias in the analysis."
),
html.Li(
"Quality of Matching: After matching, a significant overlap with similar shapes in the distributions "
"suggests successful balancing of covariates between the groups. Poor overlap or disparate shapes may "
"indicate inadequate matching."
),
html.Li(
"Potential Issues: Sparse regions in the distribution suggest a lack of overlap, meaning some treated "
"units lack suitable control matches. Additionally, outliers or extreme scores may signal issues with "
"the model or covariates."
),
]),
html.P(
"By examining these distributions, you can assess the effectiveness of the matching process and identify "
"any potential issues that could affect the reliability of the analysis."
)
],
width=4,
style={
'backgroundColor': 'rgba(173, 216, 230, 0.8)', # Pale blue color
'padding': '20px',
'border': '1px solid #ccc'
})
],
style={
'backgroundColor': 'rgba(173, 216, 230, 0.5)', # Pale blue color
'padding': '20px',
'border': '1px solid #ccc'
}),
dbc.Row([
html.H3("Balance of Covariates"),
dbc.Col([
html.H4("Explanation"),
dcc.Markdown('''
**Balance of Covariates Distribution Plots** are visual tools used to compare the distributions of covariates (variables) between treated and control groups after matching. The goal is to assess if the matching process has made the treated and control groups similar in terms of the covariates. If the matching is successful, the distributions for the treated and control groups should appear more similar, suggesting that the groups are well balanced.
**Standard Mean Difference (SMD)** is a numerical metric used to quantify the balance of covariates between treated and control groups. It measures the difference in means of a covariate between the two groups, standardised by the pooled standard deviation. An SMD close to 0 indicates that the covariate is well-balanced between the groups, while a larger SMD suggests imbalance.
**Interpreting the Plots and SMD:**
- **Well-Matched Covariates:** After matching, the covariate distributions between treated and control groups should overlap significantly, and the SMD should be close to 0.
- **Poor Matching:** If the distributions differ significantly or the SMD is still large after matching, it indicates that the matching did not adequately balance the groups.
- **SMD Threshold:** As a general rule, an SMD below 0.1 is considered acceptable, indicating a good balance between the groups.
These tools help assess the effectiveness of the matching process and ensure that the treatment effect is estimated from comparable groups.
'''),
], width=3,
style={
'backgroundColor': 'rgba(172, 230, 174, 0.8)',
'padding': '20px',
'border': '1px solid #ccc'
}),
dbc.Col([
html.H4("Plot & Standard Mean Difference (SMD)"),
html.Div([
html.Div(id='output-covariate-balance',
style={
'display': 'grid',
'gridTemplateColumns': 'repeat(auto-fill, minmax(400px, 1fr))',
'gap': '20px'})
], style={'padding': '20px'})
], width=9)
],
style={
'backgroundColor': 'rgba(172, 230, 174, 0.5)', # Pale blue color
'padding': '20px',
'border': '1px solid #ccc'
}),
dbc.Row([
html.H3("t-Test"),
dbc.Col([
html.H4("Results"),
html.Br(),
html.Div(id="ttest-output")
], width=3),
dbc.Col([
html.H4("Explanation"),
dcc.Markdown('''
In Propensity Score Matching (PSM), the **t-test** is used to assess the balance of covariates between the treated group (those who received the treatment) and the control group (those who did not receive the treatment).
The t-test evaluates the null hypothesis that the means of a specific covariate are equal between the treated and control groups. Essentially, it checks whether the difference in means is statistically significant. A well-balanced covariate between the groups is indicated by a high p-value, suggesting that the matching process has been effective.
###### How to Interpret the t-Test in PSM?
- **P-value**:
- **High P-value (e.g., p > 0.05)**: Indicates no statistically significant difference in means between the treated and control groups, suggesting that the covariate is well balanced after matching.
- **Low P-value (e.g., p < 0.05)**: Indicates a significant difference in means, suggesting that the covariate is not well balanced, even after matching.
###### Why is the t-Test Important in PSM?
The t-test is crucial for assessing the success of the matching process. By checking for balance in covariates, it helps ensure that the treated and control groups are comparable. This balance is essential for reducing confounding bias and obtaining accurate estimates of the treatment effect. Significant differences detected by the t-test may indicate that the matching process needs improvement to achieve better balance.
In summary, the t-test helps validate the effectiveness of the matching process by evaluating whether covariates are balanced between the treated and control groups, ensuring a more accurate assessment of treatment effects.
''')
], width=9,
style={
'backgroundColor': 'rgba(157, 39, 245, 0.3)',
'padding': '20px',
'border': '1px solid #ccc'
})
],
style={
'backgroundColor': 'rgba(157, 39, 245, 0.2)',
'padding': '20px',
'border': '1px solid #ccc'
})
])
#%% App Layout
app.layout = html.Div(
[
dcc.Location(id="url", refresh=False),
navbar,
dbc.Container(id="page-content", className="mb-4", fluid=True),
dcc.Store(id='stored-data', storage_type='session')
]
)
@app.callback(Output("page-content", "children"), Input("url", "pathname"))
def display_page(pathname):
if pathname == "/":
return overview_layout
elif pathname == "/match":
return match_layout
elif pathname == "/quality":
return quality_layout
else:
return dbc.Jumbotron(
[
html.H1("404: Not found", className="text-danger"),
html.Hr(),
html.P(f"The pathname {pathname} was not recognized..."),
]
)
#%% Retain Data Across Pages
#Callback to store input values automatically when they change
@app.callback(
Output('stored-data', 'data'),
Input('treatment-toggle', 'value'),
Input('upload-data', 'contents'),
Input('upload-data', 'filename'),
Input('area-id-column', 'value'),
Input('treatment-column', 'value'),
Input('covariate-columns', 'value')
)
def store_inputs(treatment_toggle, contents, filename, area_id_col, treatment_col, covariates_cols):
return {'treatment_toggle': treatment_toggle,
'contents': contents,
'filename': filename,
'area_id_col': area_id_col,
'treatment_col': treatment_col,
'covariate-columns': covariates_cols}
#Load Data When Revisit Matching
@app.callback(
Output('treatment-toggle', 'value'),
Output('upload-data', 'contents'),
Output('upload-data', 'filename'),
Output('area-id-column', 'value'),
Output('treatment-column', 'value'),
Output('covariate-columns', 'value'),
Output('run-analysis', 'n_clicks'),
Input('url', 'pathname'),
State('stored-data', 'data')
)
def restore_toggle_state(pathname, stored_data):
if pathname == '/match' and stored_data is not None:
return stored_data.get('treatment_toggle'), stored_data.get('contents'), stored_data.get('filename'), stored_data.get('area_id_col'), stored_data.get('treatment_col'), stored_data.get('covariate-columns'), 1
return dash.no_update, dash.no_update, dash.no_update, dash.no_update, dash.no_update, dash.no_update, dash.no_update
#Add Clear Button
@app.callback(
Output('stored-data', 'data', allow_duplicate=True),
Output('treatment-toggle', 'value', allow_duplicate=True),
Output('upload-data', 'contents', allow_duplicate=True),
Output('upload-data', 'filename', allow_duplicate=True),
Output('area-id-column', 'value', allow_duplicate=True),
Output('treatment-column', 'value', allow_duplicate=True),
Output('covariate-columns', 'value', allow_duplicate=True),
Output('run-analysis', 'n_clicks', allow_duplicate=True),
Input('clear-button', 'n_clicks'),
prevent_initial_call=True
)
def clear_stored_data(n_clicks):
if n_clicks:
return None, None, None, None, None, None, None, 0
return dash.no_update
#%% App Functions
# Utility function to parse the uploaded CSV file
def parse_contents(contents, filename):
content_type, content_string = contents.split(',')
decoded = base64.b64decode(content_string)
try:
if 'csv' in filename:
df = pd.read_csv(io.StringIO(decoded.decode('utf-8')))
else:
return None
except Exception as e:
print(e)
return None
return df
# Update the progress bar and dropdown options based on uploaded file
@app.callback(
Output('upload-progress', 'value'),
Output('upload-progress', 'children'),
Output('upload-progress', 'style'),
Output('area-id-column', 'options'),
Output('treatment-column', 'options'),
Output('covariate-columns', 'options'),
Input('upload-data', 'contents'),
State('upload-data', 'filename')
)
def update_progress_bar(contents, filename):
if contents is None:
return 0, "", {'backgroundColor': 'lightgray'}, [], [], []
# Start the progress bar at 50% when the file is detected
progress_value = 50
progress_text = "Uploading..."
# Parse the file
df = parse_contents(contents, filename)
if df is not None:
columns = [{'label': col, 'value': col} for col in df.columns]
# Finish the progress bar when the file is fully loaded
return 100, "Upload Complete", {'backgroundColor': 'green'}, columns, columns, columns
else:
return 0, "Failed to upload", {'backgroundColor': 'red'}, [], [], []
# Conditional analysis based on treatment selection
@app.callback(
[Output('analysis-output', 'children'),
Output('raw-results', 'data')],
Input('run-analysis', 'n_clicks'),
State('treatment-toggle', 'value'),
State('upload-data', 'contents'),
State('upload-data', 'filename'),
State('area-id-column', 'value'),
State('treatment-column', 'value'),
State('covariate-columns', 'value'),
prevent_initial_call=True
)
def run_analysis(n_clicks, treatment_toggle, contents, filename, area_id_col, treatment_col, covariates_cols):
if n_clicks == 0 or contents is None:
return "", ""
df = parse_contents(contents, filename)
if df is not None and area_id_col is not None and covariates_cols:
# Set the area ID as the index for easier referencing
df.set_index(area_id_col, inplace=True)
#Standardize Covariates
scaler = StandardScaler()
covariates_scaled = scaler.fit_transform(df[covariates_cols])
if treatment_toggle:
# Treatment areas selected: Perform propensity score matching
if treatment_col is None:
return "Please select a treatment column.", ""
if area_id_col == treatment_col:
return "The treatment column must be different from the label column.", ""
if not set(df[treatment_col].unique()).issubset({0, 1}):
return "Treatment column must contain a binary response. 1 = Treatment // 0 = Control.", ""
#Estimate Propensity Scores
logistic = LogisticRegression()
logistic.fit(covariates_scaled, df[treatment_col])
df['propensity_score'] = logistic.predict_proba(covariates_scaled)[:, 1]
#Extract Treatment & Control
treated = df[df[treatment_col] == 1]
control = df[df[treatment_col] == 0]
#Perform Matching
nn = NearestNeighbors(n_neighbors=1)
nn.fit(control[['propensity_score']])
treated_indices = df[df[treatment_col] == 1].index
control_indices = df[df[treatment_col] == 0].index
if len(treated_indices) > 0:
distances, indices = nn.kneighbors(treated[['propensity_score']])
matched_data = pd.DataFrame({
'Treated_Area': treated_indices,
'Matched_Control_Area': control.iloc[indices.flatten()].index,
'Distance': distances.flatten()
})
# Create raw results dataframe
raw_results = matched_data.copy()
return dbc.Table.from_dataframe(matched_data, striped=True, bordered=True, hover=True), raw_results.to_dict('records')
return "No treated areas to match.", ""
else:
# No treatment: Find similar areas using nearest neighbors
nn = NearestNeighbors(n_neighbors=2) # 2 to get the nearest neighbor excluding the point itself
nn.fit(covariates_scaled)
distances, indices = nn.kneighbors(covariates_scaled)
#Filter Self Returns
indicesResults = []
for i in range(len(indices)):
#If NN 1 == index select NN 2 else NN 1
if int(indices[i][0]) == i:
indicesResults.append((indices[i][1], distances[i][1]))
else:
indicesResults.append((indices[i][0], distances[i][0]))
indicesResults = np.array(indicesResults)
similar_areas = pd.DataFrame({
'Area': df.index,
'Most_Similar_Area': df.index[indicesResults[:, 0].astype(int)],
'Similarity_Score': indicesResults[:, 1]
})
# Create raw results dataframe
raw_results = similar_areas.copy()
return dbc.Table.from_dataframe(similar_areas, striped=True, bordered=True, hover=True), raw_results.to_dict('records')
return "Please upload a valid CSV file and select the appropriate columns.", ""
# Example callback for downloading data
@app.callback(
Output("download-data", "data"),
Input("download-button", "n_clicks"),
State("raw-results", "data"),
prevent_initial_call=True
)
def download_results(n_clicks, raw_results):
if n_clicks > 0 and raw_results is not None:
# Convert raw results to DataFrame and then to CSV
df_raw = pd.DataFrame(raw_results)
# Get the current time
now = datetime.now()
current_dt = now.strftime("%Y-%m-%d_%H%M%S")
return dcc.send_data_frame(df_raw.to_csv, "MOPAC_SpatialMatching_"+str(current_dt)+".csv")
#%% Quality Assurance Functions
def calculate_smd(df, treatment_col, covariates_cols, state):
smd = []
treated = df[df[treatment_col] == 1]
control = df[df[treatment_col] == 0]
for cov in covariates_cols:
treated_mean = treated[cov].mean()
control_mean = control[cov].mean()
treated_std = treated[cov].std()
control_std = control[cov].std()
pooled_std = np.sqrt((treated_std**2 + control_std**2) / 2)
#smd[cov] = np.abs(treated_mean - control_mean) / pooled_std
smd.append(f'SMD {state}: {np.abs(treated_mean - control_mean) / pooled_std}')
return smd
def plot_covariate_distributions(df, treatment_col, covariates_cols):
plots = []
for cov in covariates_cols:
fig = go.Figure()
# Define colors for treatments
treatments = df[treatment_col].unique()
colors = ['blue', 'green'] # Customize as needed
treatment_colors = {treatment: colors[i % len(colors)] for i, treatment in enumerate(treatments)}
label_mapping = {1: 'Treatment', 0: 'Control'} # Adjust as needed
# Add a histogram trace for each treatment
for treatment in treatments:
subset = df[df[treatment_col] == treatment]
color = treatment_colors[treatment]
# Add histogram trace
fig.add_trace(go.Histogram(
x=subset[cov],
name=label_mapping.get(treatment, str(treatment)),
opacity=0.6,
histnorm='probability density',
nbinsx=30,
bingroup=1, # Ensure histograms are layered
marker_color=color # Set color of histogram bars
))
# Calculate KDE for each subset
kde = gaussian_kde(subset[cov].dropna())
x = np.linspace(subset[cov].min(), subset[cov].max(), 1000)
kde_y = kde(x)
# Add KDE trace with the same color
fig.add_trace(go.Scatter(
x=x,
y=kde_y,
mode='lines',
name=f'KDE {label_mapping.get(treatment, str(treatment))}',
line=dict(color=color, width=2), # Set color of KDE line
showlegend=False # Hide KDE lines from the legend
))
# Update layout
fig.update_layout(
title=f'Distribution of {cov}',
xaxis_title=cov,
yaxis_title='Density',
barmode='overlay', # Layer the histograms
legend_title='Treatment',
bargap=0.1
)
plots.append(fig)
return plots
def plot_propensity_scores(df, treatment_col):
fig = go.Figure()
# Define labels for treatment and control
label_mapping = {1: 'Treatment', 0: 'Control'}
# Get unique treatments
treatments = df[treatment_col].unique()
# Define a color map (you can customize these colors)
colors = ['blue', 'green']
# Ensure the number of colors matches the number of treatments
treatment_colors = {treatment: colors[i % len(colors)] for i, treatment in enumerate(treatments)}
# Add a histogram trace for each treatment
for treatment in treatments:
subset = df[df[treatment_col] == treatment]
color = treatment_colors[treatment]
# Add histogram trace
fig.add_trace(go.Histogram(
x=subset['propensity_score'],
name=label_mapping.get(treatment, str(treatment)), # Set custom label
opacity=0.6,
histnorm='probability density', # Use density
nbinsx=30,
bingroup=1, # Ensures histograms are layered
marker_color=color # Set the color of the histogram bars
))
# Calculate KDE for each subset
kde = gaussian_kde(subset['propensity_score'].dropna())
x = np.linspace(subset['propensity_score'].min(), subset['propensity_score'].max(), 1000)
kde_y = kde(x)
# Add KDE trace with the same color
fig.add_trace(go.Scatter(
x=x,
y=kde_y,
mode='lines',
name=f'KDE {label_mapping.get(treatment, str(treatment))}', # Set custom label for KDE
line=dict(color=color, width=2), # Set the color of the KDE line
showlegend=False # Hide KDE lines from the legend
))
# Update layout
fig.update_layout(
title='Distribution of Propensity Scores',
xaxis_title='Propensity Score',
yaxis_title='Density', # Use 'Density' on the Y-axis
barmode='overlay', # Layer the histograms
legend_title='Group',
bargap=0.1
)
return fig
def t_test_covariates(df, treatment_col, covariates_cols):
t_test_results = {}
treated = df[df[treatment_col] == 1]
control = df[df[treatment_col] == 0]
for cov in covariates_cols:
t_stat, p_value = stats.ttest_ind(treated[cov], control[cov], equal_var=False)
t_test_results[cov] = {'t_stat': t_stat, 'p_value': p_value}
return t_test_results
@app.callback(
Output('output-quality-error', 'children'),
Output('output-covariate-balance', 'children'),
Output('ttest-output', 'children'),
Output('propensity-distribution', 'children'),
Input('run-quality', 'n_clicks'),
Input('url', 'pathname'),
Input('stored-data', 'data')
)
def quality_assurance(n_clicks, pathname, stored_data):
if pathname == '/quality' and n_clicks:
if stored_data is None:
return "Error: Please Run the Analysis First.", "", "", ""
#Get the data
treatment_toggle = stored_data.get('treatment_toggle')
file_contents = stored_data.get('contents')
file_name = stored_data.get('filename')
area_id_col = stored_data.get('area_id_col')
treatment_col = stored_data.get('treatment_col')
covariate_cols = stored_data.get('covariate-columns')
if any(value is None for value in stored_data.values()):
return "Error: Please Run the Treatment-Control Matching First.", "", "", ""
if not treatment_toggle:
return "Error: Quality Assurance Requires a Treatment and Control Group", "", "", ""
if treatment_col is None:
return "Error: Please select a treatment column.", "", "", ""
if area_id_col == treatment_col:
return "Error: The treatment column must be different from the label column.", "", "", ""
#Parse data
df = parse_contents(file_contents, file_name)
if not set(df[treatment_col].unique()).issubset({0, 1}):
return "Error: Treatment column must contain a binary response. 1 = Treatment // 0 = Control.", "", "", ""
#Check Treated Areas True
if len(df[df[treatment_col] == 1].index) == 0:
return "Error: No treated areas to match.", "", "", ""
def generate_match_data(df, treatment_col, covariate_cols):
#Get subset of data
treatment = df[treatment_col]
covariates = df[covariate_cols]
#Data Transform
scaler = StandardScaler()
covariates_scaled = scaler.fit_transform(covariates)
#Propensity Scores
logistic = LogisticRegression()
logistic.fit(covariates_scaled, treatment)
# Get the propensity scores
df['propensity_score'] = logistic.predict_proba(covariates_scaled)[:, 1]
#Subset data
treated = df[df[treatment_col] == 1]
control = df[df[treatment_col] == 0]
#NN Matching
nn = NearestNeighbors(n_neighbors=1)
nn.fit(control[['propensity_score']])
distances, indices = nn.kneighbors(treated[['propensity_score']])
matched_control = control.iloc[indices.flatten()]
df_matched = pd.concat([treated, matched_control])
return(df_matched)
df_matched = generate_match_data(df, treatment_col, covariate_cols)
## BALANCE OF COVARIATES
plots_boc = plot_covariate_distributions(df_matched, treatment_col, covariate_cols)
stats_boc_b = calculate_smd(df, treatment_col, covariate_cols, 'Before')
stats_boc_a = calculate_smd(df_matched, treatment_col, covariate_cols, 'After')
out_covariate_balance = []
for plot, stat_b, stat_a in zip(plots_boc, stats_boc_b, stats_boc_a):
out_covariate_balance.append(html.Div([dcc.Graph(figure=plot),
html.Br(),
html.P(stat_b),
html.P(stat_a)], className='grid-item'))
## Propensity Distribution
prop_plot = plot_propensity_scores(df, treatment_col)
## t-TEST RESULTS
t_test_results = t_test_covariates(df, treatment_col, covariate_cols)
out_t_test = []
for main_key, sub_dict in t_test_results.items():
p_value = sub_dict.get('p_value')
t_stat = sub_dict.get('t_stat')
out_t_test.append(html.Div([html.H6(main_key),
html.Ul([
html.Li(f'p-Value: {p_value}'),
html.Li(f't-Stat: {t_stat}')
])
]))
return "", out_covariate_balance, out_t_test, html.Div([dcc.Graph(figure=prop_plot)])
else:
return "", "", "", ""
#%% Main
# Run the app
if __name__ == '__main__':
app.run_server(debug=False, host='0.0.0.0', port=int(os.environ.get('PORT', 7860)))
|