samplics 0.0.9

Select, weight and analyze complex sample data

samplics is a python package for selecting, weighting and analyzing sample obtained from complex sampling design.

Sample Analytics

In large scale surveys, often complex random mechanisms are used to select samples. Estimations obtained from such samples must reflect the random mechanism to ensure accurate calculations. samplics implements a set of sampling techniques for complex survey designs.

Selection

Since the full population cannot be observed, a sample is selected to estimate population parameters of interest. The assumption is that the sample is representative of the population for the characteristics of interest. The selection methods in samplics are:

• Simple random sampling (SRS)

• Systematic selection (SYS)

• Probability proportional to size (PPS)

  • Systematic

  • Brewer’s method

  • Hanurav-Vijayan method

  • Murphy’s method

  • Sampford’s method

• Unequal sample selection

Weighting

Sample weighting is the main mechanism used in surveys to formalize the representivity of the sample. The base or design weights are usually adjusted to compensate for distorsions due nonresponse and other shorcomings of the operationalization of the sampling design.

• Weight adjustment due to nonresponse

• Weight poststratification, calibration and normalization

• Weight replication i.e. Boostrap, BRR, and Jackknife

Estimation

The estimation of the parameters of interest must reflect the sampling mechanism and the weight adjustments.

• Taylor linearization procedures

• Replicate-based estimation i.e. Boostrap, BRR, and Jackknife

• Regression-based

Parameters of interest * Linear parameters e.g. total, mean, proportion * Non-linear (complex) parameters e.g. ratio, regression coefficient

Installation

pip install samplics

if both Python 2.x and python 3.x is installed on your computer, you may have to use: pip3 install samplics

Dependencies

Python versions 3.6.x or newer and the following packages:

• numpy

• pandas

• scpy

• statsmodels

Usage

To select a sample of primary sampling units using PPS method, we can use a code similar to:

import samplics
from samplics.sampling import Sample

psu_frame = pd.read_csv("psu_frame.csv")
psu_sample_size = {"East":3, "West": 2, "North": 2, "South": 3}
pps_design = Sample(method="pps-sys", stratification=True, with_replacement=False)
frame["psu_prob"] = pps_design.inclusion_probs(
    psu_frame["cluster"],
    psu_sample_size,
    psu_frame["region"],
    psu_frame["number_households_census"]
    )

To adjust the design sample weight for nonresponse, we can use a code similar to:

import samplics
from samplics.weighting import SampleWeight

status_mapping = {
    "in": "ineligible", "rr": "respondent", "nr": "non-respondent", "uk":"unknown"
    }

full_sample["nr_weight"] = SampleWeight().adjust(
    samp_weight=full_sample["design_weight"],
    adjust_class=full_sample["region"],
    resp_status=full_sample["response_status"],
    resp_dict=status_mapping
    )

import samplics
from samplics.estimation import TaylorEstimation, ReplicateEstimator

zinc_mean_str = TaylorEstimator("mean").estimate(
    y=nhanes2f["zinc"],
    samp_weight=nhanes2f["finalwgt"],
    stratum=nhanes2f["stratid"],
    psu=nhanes2f["psuid"],
    exclude_nan=True
)

ratio_wgt_hgt = ReplicateEstimator("brr", "ratio").estimate(
    y=nhanes2brr["weight"],
    samp_weight=nhanes2brr["finalwgt"],
    x=nhanes2brr["height"],
    rep_weights=nhanes2brr.loc[:, "brr_1":"brr_32"],
    exclude_nan = True
)

Contributing

TBD

License

MIT

Project status

This is an alpha version. At this stage, this project is not recommended to be used for production or any project that the user depend on.