epiinfo 1.2.0.17

Epi Info: Import and analyze data

Epi Info Data Imports and Analyses

This package brings Epi Info statistical analysis routines, plus a log binomial regression analysis routine for computing adjusted relative risks of associations with binary outcome data, to the Python environment. It has methods for converting datasets into Python objects and for passing those objects to analysis functions.

Install with pip

pip install epiinfo

Full path to pip may be necessary. If so, try the Scripts subdirectory of the Python install directory.

Classes

ImportData

Contains functions for importing data sources to Python objects (lists of dictionaries) for consumption by analysis functions.

from epiinfo.ImportData import *

CSV

dataset = eicsv('path/filename.csv')

Reads a CSV file and creates a list of dictionaries called 'dataset'.

JSON

dataset = eijson('path/filename.json')

Reads a JSON file and creates a list of dictionaries called 'dataset'.

SQLite

datadict = {}
dataset = eisqlite('path/filename.db', 'tablename', datadict)

Reads a table from a(n) SQLite database and creates a list of dictionaries called 'dataset'.
Also populates an initially-empty dict with columns (keys) and SQLite data types (values).

Sync File

initVector = 'D1041B94D49F66DF120AD40269E102EA'
passwordSalt = '15A4D4258940BD492EE2'
dataset = eisync('path/filename.epi7', initVector, passwordSalt, 'password')

Reads a sync file (encrypted XML exported from Epi Info mobile apps) and creates a list of dictionaries called 'dataset'. initVector is the Init Vector that was generated by the mobile device to encrypt the data on the mobile device. passwordSalt is the Salt that was generated by the mobile device to encrypt the data on the mobile device. password is the password that was used to encrypt the data on the mobile device.

This function can only be use with data that was encrypted using the Epi Info iOS app's Custom Keys option. See here for details.

MakeSQLite

Contains a function for writing a list of dict data object to a SQLite table.

from epiinfo.MakeSQLite import *
makesql = MakeSQLite()
res = makesql.makeTable('C:\\Path\\to\\SQLite\\Database\\File\\sql1.db', dataobject, 'sqltable', sqldatatypes)

• The first argument to the makeTable function is the path and file name for the SQLite database to which to write or replace the table.
  
• dataobject is the Python list of dictionaries to be written to SQLite.
  
• 'sqltable' is the name of the destination SQLite table.
  
• sqldatatypes is a dict of column names and SQLite data types to be used in creating the SQLite table. This dict, by this name, is automatically created by the Epi Info Classic Analysis PYTHON statement.

TablesAnalysis

Performs MxN Tables Analyses of Outcome and Exposures, with statistical tests of association.

from epiinfo.TablesAnalysis import *
ivdict = {'outcomeVariable' : 'Ill'}
ivdict['exposureVariables'] = ['ChefSalad', 'Workdays']
ta = TablesAnalysis()
taResults = ta.Run(ivdict, dataset)

'dataset' is the list of dictionaries containing the analysis data. 'ivdict' is a dictionary of analysis variables.

The result of Run() is this dict:

{'Variables': ['Ill * ChefSalad', 'Ill * Workdays'],
 'VariableValues': [[[1, 0], [True, False]],
  [[1, 0], [0, 1, 2, 3, 4, 5, 6, 7]]],
 'Tables': [[[150, 79], [43, 37]],
  [[7, 4],
   [21, 10],
   [21, 12],
   [29, 16],
   [18, 15],
   [25, 9],
   [15, 12],
   [57, 38]]],
 'RowTotals': [[229, 80], [11, 31, 33, 45, 33, 34, 27, 95]],
 'ColumnTotals': [[193, 116], [193, 116]],
 'Statistics': [{'OR': 1.633794524580512,
   'ORLL': 0.974133414274391,
   'ORUL': 2.740163215258917,
   'MidPOR': 1.6311162097465466,
   'MidPORLL': 0.9691190383377568,
   'MidPORUL': 2.7422471723180135,
   'FisherORLL': 0.9399306300659928,
   'FisherORUL': 2.826676338762746,
   'MidPExact1Tail': 0.032696255288936306,
   'FisherExact1Tail': 0.04214460217937938,
   'FisherExact2Tail': 0.08083332652929953,
   'RiskRatio': 1.2186452726718797,
   'RiskRatioLL': 0.9741280246023387,
   'RiskRatioUL': 1.5245391397211574,
   'RiskDifference': 11.752183406113536,
   'RiskDifferenceLL': -0.789033558172958,
   'RiskDifferenceUL': 24.293400370400033,
   'UncorrectedX2': 3.4922535822177045,
   'UncorrectedX2P': 0.0617,
   'MHX2': 3.4809517906894913,
   'MHX2P': 0.0621,
   'CorrectedX2': 3.0090265314124536,
   'CorrectedX2P': 0.0828},
  {'ChiSq': 3.922874010476977,
   'ChiSqDF': 7,
   'ChiSqP': 0.7886,
   'FishersExact': 0.7683844417832235}]}

Results are from the fictional Salmonellosis dataset often used in Epi Info training sessions, with the 8-value variable "Workdays" added to demonstrate 2xN analysis.

(MxN Fisher's Exact test is adapted from the 1983 Fortran (FORTRAN 77) routine by Mehta and Patel.)

LogisticRegression

Contains functions for Logistic Regression analysis. In addition to computing identical results to Epi Info Dashboard on identical datasets, this routine also computes interaction odds ratios and confidents limits for interaction effects that are not also main effects.

from epiinfo.LogisticRegression import *
ivdict = {'Ill' : 'dependvar'} #, 'intercept' : True , 'includemissing' : False} # intercept and includemissing are optional keys
ivdict['exposureVariables'] = ['column0', 'column1', column2, 'column0*column1']
# ivdict['Group'] = 'matchgroupcolumn' # optional for matched case control studies
reg = LogisticRegression()
rslts = reg.doRegression(ivdict, dataset)

'dataset' is the list of dictionaries containing the analysis data. 'ivdict' is a dictionary of analysis variables and options.

rslts.show()

LOGISTIC REGRESSION RESULTS
Variable          Coefficient    Standard Error    Odds Ratio    Lower   Upper  
ChefSalad            1.145           0.3429          3.1424      1.6046  6.1539 
EggSaladSandwich     1.0418          0.3146          2.8343       1.53   5.2506 
CONSTANT            -0.7644          0.3602     

Number of Iterations             4
Minus Two Log-Likelihood    393.37
Number of Observations         309

Fit Test             Value      DF       P
Score              14.7777       2  0.0006
Likelihood Ratio   15.5999       2  0.0004

Individual result functions provide greater precision, as well as meaningful odds ratios for interaction terms.

print(rslts.Variables)
print(rslts.Beta)
print(rslts.SE)
print(rslts.OR)
print(rslts.ORLCL)
print(rslts.ORUCL)
print(rslts.Z)
print(rslts.PZ)
print(rslts.Iterations)
print(rslts.MinusTwoLogLikelihood)
print(rslts.CasesIncluded)
print(rslts.Score, rslts.ScoreDF, rslts.ScoreP)
print(rslts.LikelihoodRatio, rslts.LikelihoodRatioDF, rslts.LikelihoodRatioP)
for ior in rslts.InteractionOR:
    print(ior)

['ChefSalad', 'EggSaladSandwich', 'CONSTANT']
[1.1449910849486071, 1.041809442441627, -0.7644353230444337]
[0.34290905726619875, 0.31456430262135265, 0.3601766334566801]
[3.1424133419378486, 2.834340954032848]
[1.6046436032972575, 1.5300107079664744]
[6.153865937145245, 5.250609425070682]
[3.3390517418143193, 3.311912488988536, -2.1223901053991483]
[0.0008406490630090172, 0.0009266052827040479, 0.03380499507237897]
4
393.3736390464802
309
14.777703590233484 2 0.0006
15.599932822362803 2 0.0004

Results are from the fictional Salmonellosis dataset often used in Epi Info training sessions.

rslts.showHTML()

Formats rslts.show() as HTML for the Output pane within Epi Info Classic Analysis.

LogBinomialRegression

Contains functions for Log-Binomial Regression analysis to provide adjusted risk ratios when the independent variable is binary.

from epiinfo.LogBinomialRegression import *
ivdict = {'Ill' : 'dependvar'} #, 'intercept' : True , 'includemissing' : False} # intercept and includemissing are optional keys
ivdict['exposureVariables'] = ['column0', 'column1', column2, 'column0*column1']
#ivl['StartValues'] = [0.0, 0.0, 0.0, 0.0, -0.45] # Optional user-supplied starting beta values
reg = LogBinomialRegression()
rslts = reg.doRegression(ivdict, dataset)

'dataset' is the list of dictionaries containing the analysis data. 'ivdict' is a dictionary of analysis variables and options.

rslts.show()

LOG BINOMIAL REGRESSION RESULTS
Variable          Coefficient    Standard Error    Risk Ratio    Lower   Upper  
EggSaladSandwich     0.287           0.0863          1.3325      1.1251  1.578  
ChefSalad            0.3359          0.1156          1.3992      1.1155  1.755  
CONSTANT            -0.8579          0.1275     

Number of Iterations             7
Log-Likelihood             -197.81
Number of Observations         309

Individual result functions provide greater precision, as well as parameter iteration history and meaningful risk ratios for interaction terms.

print(rslts.Variables)
print(rslts.Beta)
print(rslts.SE)
print(rslts.OR)
print(rslts.ORLCL)
print(rslts.ORUCL)
print(rslts.Z)
print(rslts.PZ)
print(rslts.Iterations)
print(rslts.MinusTwoLogLikelihood)
print(rslts.CasesIncluded)
for ior in rslts.InteractionOR:
    print(ior)
# To see start values and subsequent beta parameter values in the iteration history.
print()
for ph in rslts.ParameterHistory:
    print(ph)

['EggSaladSandwich', 'ChefSalad', 'CONSTANT']
[0.28702656278093225, 0.33587378874771817, -0.8578882747952187]
[0.08629620218172233, 0.11560278700279132, 0.127540265427839]
[1.3324596068382382, 1.399162423625005]
[1.1251193436636948, 1.1154930317062357]
[1.5780091364122923, 1.7549688183079077]
[3.3260625094081475, 2.90541255497247, -6.726411238971455]
[0.0008808217459920762, 0.003667693292066409, 1.7389864495598986e-11]
7
-197.80804698840117
309

[0.015412272950627526, 0.018694620739870826, -0.1434745513345579]
[0.047175480510386114, 0.053313801454389464, -0.2716280030939807]
[0.12025288102095905, 0.13274826883584623, -0.4755035906059664]
[0.22117400493838188, 0.24761248723857526, -0.7059060377582919]
[0.2784656944889591, 0.3214894897321413, -0.8352770236314679]
[0.2868833735148343, 0.33549848896238826, -0.857374553660365]
[0.28702656278093225, 0.33587378874771817, -0.8578882747952187]

Results are from the fictional Salmonellosis dataset often used in Epi Info training sessions.

rslts.showHTML()

Formats rslts.show() as HTML for the Output pane within Epi Info Classic Analysis.

EICSTables

Contains the ComplexSampleTables class, which contains the functions ComplexSampleTables and ComplexSampleFrequencies.

from epiinfo.EICSTables import *
csTables = ComplexSampleTables()
results = csTables.ComplexSampleTables(ivdict, dataset)
results = csTables.ComplexSampleFrequencies(ivdict, dataset)

'dataset' is the list of dictionaries containing the analysis data. 'ivdict' is a dictionary of analysis variables.

EICSMeans

Contains the ComplexSampleMeans class, which contains the ComplexSampleMeans function.

from epiinfo.EICSMeans import *
csMeans = ComplexSampleMeans()
results = csMeans.ComplexSampleMeans(ivdict, dataset)

'dataset' is the list of dictionaries containing the analysis data. 'ivdict' is a dictionary of analysis variables.