Useful package to help download and merge data from NHANES medical database. Can clean/filter data, do statistical comparison analysis, and perform multivariate logistic regression.

Example downloading and cleaning data:

from nhutils.generate import create_dataset
from nhutils.clean import Scrubber

# define parameters of interest for new dataset
years_of_interest = ['2011-2012', '2013-2014', '2015-2016', '2017-2018']
var_names = ['SEQN', 'RIDAGEYR', 'RIAGENDR', 'DMDEDUC2', 'DIQ080', 'SMQ020']
# download and create dataset
df = create_dataset(vars=var_names, years=years_of_interest)
# clean data by using Scrubber class
scrub = Scrubber(df)
scrub.minus_one('RIAGENDR')
scrub.remove_7_and_9('DMDEDUC2')
scrub.convert_to_binary(['DIQ080', 'SMQ020'])
# save cleansed data to csv
df.to_csv('dataset.csv', index=False)

Example of comparison anaylsis:

import pandas as pd
from nhutils.stats import compare_stats

# import data
data = pd.read_excel('dataset.xlsx')
# define groups to be compared
dr = data[data['DIQ080'] == 1]
diab = data[(data['DIQ010'] == 1) & (data['DIQ080'] == 0)]
# call compare_stats function and provide function parameters
results = compare_stats(group1=dr, 
                         group1_label='DR', 
                         group2=diab, 
                         group2_label='Diabetes',
                         numerical_vars=['RIDAGEYR', 'URDACT', 'BMXBMI'],
                         categorical_vars=['RIAGENDR', 'RIDRETH1', 'DMDEDUC2'],
                         output_excel_filename='DRvsDiabetes.xlsx',
                         welchs_t_test=True,
                         decimal_places=4)

After running above code block, results would look like this:

variable	DR	Diabetes	p-value
RIDAGEYR	62.9549 (12.424)	61.2482 (13.9428)	0.0022
URDACT	313.9478 (969.5149)	136.8675 (613.404)	0
BMXBMI	32.3415 (8.2252)	32.3103 (7.5454)	0.9321
RIAGENDR			0.0823
RIAGENDR = 0	55.3383	51.4528
RIAGENDR = 1	44.6617	48.5472
RIDRETH1			0.1905
RIDRETH1 = 1	16.391	16.6667
RIDRETH1 = 2	12.0301	9.9677
RIDRETH1 = 3	28.8722	32.6069
RIDRETH1 = 4	26.6165	26.7958
RIDRETH1 = 5	16.0902	13.9629
DMDEDUC2			0.0727
DMDEDUC2 = 1.0	18.6747	15.91
DMDEDUC2 = 2.0	16.5663	14.9284
DMDEDUC2 = 3.0	23.7952	22.4949
DMDEDUC2 = 4.0	26.3554	28.3436
DMDEDUC2 = 5.0	14.6084	18.3231

Example of multivariate logistic regression:

import pandas as pd
from nhutils.stats import log_reg

# import data
df = pd.read_excel('dataset.xlsx')
# define group to do logistic regression on
diabetes = df[df['DIQ010'] == 1]
# call log_reg function and provide function parameters
results = log_reg(data=diabetes,
                  dependent_var='depression',
                  independent_numerical_vars=['RIDAGEYR', 'BMXBMI'],
                  independent_categorical_vars=['DIQ080', 'RIAGENDR', 'RIDRETH1', 'DMDEDUC2'],
                  output_excel_filename='log_reg_results.xlsx')

After running above code block, results would look like this:

	OR	p-value	2.5%	97.5%
const	0.072711	4.85E-10	0.03185	0.165991
RIDAGEYR	0.990237	0.027291	0.981649	0.998901
BMXBMI	1.043185	8.88E-10	1.029174	1.057386
DIQ080=1.0	1.414383	0.007342	1.097718	1.822398
RIAGENDR=1	1.738442	1.11E-06	1.391635	2.171678
RIDRETH1=2	1.748111	0.005543	1.178047	2.594034
RIDRETH1=3	1.606838	0.008162	1.130741	2.283394
RIDRETH1=4	1.020704	0.912587	0.707948	1.471628
RIDRETH1=5	1.090866	0.707666	0.692392	1.718663
DMDEDUC2=2.0	0.947286	0.76531	0.663837	1.351762
DMDEDUC2=3.0	0.522351	0.000372	0.365311	0.746899
DMDEDUC2=4.0	0.518183	0.000192	0.366803	0.732039
DMDEDUC2=5.0	0.276303	3.57E-08	0.174871	0.436569