# Rwanda - Integrated Household Living Conditions Survey 2000-2001

Reference ID | RWA-NISR-EICV-2001-v1.1 |

Year | 1999 - 2001 |

Country | Rwanda |

Producer(s) | NISR (National Institute of Statistics, Rwanda) - Government of Rwanda |

Sponsor(s) | Department for Intenational Development - DFID - Bilateral funding assistance World Bank - WB - Financial assistance United Nations for the Children - UNICEF - Financial assistance United Nations for Development Program - UNDP - Financial |

Collection(s) | |

Metadata | Documentation in PDF |

Created on

Aug 03, 2012

Last modified

Jun 28, 2016

Page views

645750

Sampling

Sampling Procedure

The sampling plan was drawn up with the technical support of the late Christopher SCOTT, Survey Consultant, during his mission in July 1997.

Constraints

The two main factors considered in designing the sampling plan were:

- the objectives of the survey,

- the fieldwork methodology given the available logistical resources.

For the survey one objective was determinant: the Government wanted statistically reliable results at the level of each province, Kigali city and the “other urban sector”. Thus, the objective called for 13 domain of analysis. Experience of conducting this type of survey shows that a minimum sample of 500 households per domain of study is required for sound analyses.

Sample size

The sample size was therefore 6,450 households, with 1,170 households for urban areas and 5,280 households for rural areas.

Two stage sampling

A two stage stratified sample

was used: sampling at area level and at household level.

Sampling base

*At the area level, the chosen sampling base ( or at the enumeration district) was the “cellule”in the

rural areas and the zone in urban areas, since they are usually fairly homogeneous in size and are well

demarcated.

Knowledge of the size of each cellule enabled the use of the classical method of sampling with probability proportional to size at the first stage. A list of all cellules including estimates of the number of households in each was compiled from information provided by the local authorities.

*For sampling at the household level, an up-dated list of households was prepared for each of the selected first stage cellule by carrying out a listing in each sampled cellule simultaneously but with a lag in data collection before or while collecting the data. Part of this operation was carried out in collaboration with the National Population Office (ONAPO) and the Food Security Research Project

(FSRP) of MINAGRI.

Weighting

In order for the estimates from each survey to be representative at the national level, it is necessary to apply sampling weights to the survey data. The weights for the sample households were calculated as the inverse of the overall probability of selection, taking into account each sampling stage. Given the nature of the sample design and the new listing of households, the weights vary by sample ZD. An Excel spreadsheet with all the sampling frame information for the sample ZDs was used for calculating the weights, which were then attached to the corresponding records in the survey data files.

WEIGHTING

There are two kinds of weighting: spatial weighting and temporal weighting. Use of these methods enabled annual estimates to be obtained for the whole of the Rwandan population.

* Spatial weighting

Spatial weighting enables results relating to the sample to be extrapolated for the whole of the population for the same period. It was calculated using the inverse of the overall probability of selection of a particular household. The details of the theory for calculating the various probabilities

are shown in Annex I.Starting from the overall probability formula Fhi=p1hi x p2hi where p1hi is the probability proportional to size of drawing cellule i in stratum h and p2hi is the conditional probability of drawing a household knowing that unit i of stratum h has been selected. The numbers 1 and 2 indicate the stage or level of sampling.Spatial weighting is given by the formula Whi=1/Fhi=Mhi/ahbhi where Mhi is the total number of households in unit i of stratum h

and ah is the number of sample units in stratum h and bhi is the number of households surveyed in unit i of stratum h .

*Temporal weighting

Temporal weighting is intended to produce annual estimates of values relating to the survey period.Thus, the temporal weighting coefficient depends on the length of the collection period.By using CPTmj to designate the coefficient of temporal weighting of the variable ymj for household m, and Jmj to designate the number of collection days

Ymj=CPTmj x ymj or CPTmj=365/Jmj

Ymi being the annual value of the variable ymj for household m.