Nigeria - Multiple Indicator Cluster Survey MICS3 (2007), Nigeria, Third round

Reference ID

NGA-NBS-MICS3 2007-v1.2

Year

2007

Country

Nigeria

Producer(s)

National Bureau of Statistics [nbs] - Federal Government of Nigeria

Sponsor(s)

Fedral Government of Nigeria - FG - Funding
United Nation Children Educational Fund - UNICEF - Funding
National Bureau of Statistics - NBS - Funding

Metadata

Download DDI Download RDF

Created on

Oct 18, 2010

Last modified

Dec 02, 2013

Page views

539763

Downloads

32063

Data Appraisal

Estimates of Sampling Error

The sample of respondents selected in the Nigeria Multiple Indicator Cluster Survey is only one of the
samples that could have been selected from the same population, using the same design and size. Each of
these samples would yield results that differ somewhat from the results of the actual sample selected.
Sampling errors are a measure of the variability between all possible samples. The extent of variability is
not known exactly, but can be estimated statistically from the survey results.
The following sampling error measures are presented in this appendix for each of the selected indicators:
?? Standard error (se): Sampling errors are usually measured in terms of standard errors for particular
indicators (means, proportions etc). Standard error is the square root of the variance. The Taylor
linearization method is used for the estimation of standard errors.
?? Coefficient of variation (se/r) is the ratio of the standard error to the value of the indicator
?? Design effect (deff) is the ratio of the actual variance of an indicator, under the sampling method used
in the survey, to the variance calculated under the assumption of simple random sampling. The
square root of the design effect (deft) is used to show the efficiency of the sample design. A deft value
of 1.0 indicates that the sample design is as efficient as a simple random sample, while a deft value
above 1.0 indicates the increase in the standard error due to the use of a more complex sample design.
?? Confidence limits are calculated to show the interval within which the true value for the population
can be reasonably assumed to fall. For any given statistic calculated from the survey, the value of that
statistics will fall within a range of plus or minus two times the standard error (p + 2.se or p – 2.se) of
the statistic in 95 percent of all possible samples of identical size and design.
For the calculation of sampling errors from MICS data, SPSS Version 15 Complex Samples module has
been used. The results are shown in the tables that follow. In addition to the sampling error measures
described above, the tables also include weighted and unweighted counts of denominators for each
indicator.
Sampling errors are calculated for indicators of primary interest, for the national total, for the regions,
and for urban and rural areas. Three of the selected indicators are based on households, 8 are based on
household members, 13 are based on women, and 15 are based on children under 5. All indicators
presented here are in the form of proportions. Table SE.1 shows the list of indicators for which sampling
errors are calculated, including the base population (denominator) for each indicator. Tables SE.2 to SE.9
show the calculated sampling errors.
Table SE.1: Indicators selected for sampling error calculations
List of indicators selected for sampling error calculations, and base populations (denominators)
for each indicator, Nigeria 2007
Table SE.2: Sampling errors: Country
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE.3: Sampling errors: Urban
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Confidence limits Table
Table SE.4: Sampling errors: Rural
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE.5: Sampling errors: North East
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE6: Sampling errors: North East
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE.7: Sampling errors: South East
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE.8: Sampling errors: South South
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007
Table SE.9: Sampling errors: South West
Standard errors, coefficients of variation, design effects (deff), square root of design effects (deft) and confidence intervals for selected
indicators, Nigeria 2007

Other forms of Data Appraisal

A series of tables and graphs were genenrated
Table DQ.1: Age distribution of household population
Single-year distribution of household population by sex (weighted), Nigeria, 2007
Table DQ.2: Age distribution of eligible and interviewed women
Household population of women age 10-54, interviewed women age 15-49, and percentage of
eligible women who were interviewed (weighted), by five-year age group, Nigeria, 2007
Table DQ.3: Age distribution of eligible and interviewed under-5s
Household population of children age 0-7, children whose mothers/caretakers were interviewed
and percentage of under-5 children whose mothers/caretakers were interviewed (weighted), by
five-year age group, Nigeria, 2007
Table DQ.4: Age distribution of under-5 children
Age distribution of under-5 children by 3-month groups (weighted), Nigeria, 2007
Table DQ.5: Heaping on ages and periods
Age and period ratios at boundaries of eligibility by type of information collected (Household
questionnaire, weighted), Nigeria, 2007
Table DQ.6: Percentage of observations missing information for selected questions and indicators
(Under-5 questionnaire, weighted), Nigeria, 2007
Table DQ.7: Presence of mother in the household and the person interviewed for the under-5 questionnaire: Distribution of children under five by
whether the mother lives in the same household, and the person interviewed for the under-5 questionnaire (weighted), Nigeria, 2007
Table DQ.8: School attendance by single age
Distribution of household population age 5-24 by educational level and grade attended in the current year, Nigeria, 2007
Table DQ.9: Sex ratio at birth among children ever born and living
Sex ratio at birth among children ever born, children living, and deceased children by age of women (weighted), Nigeria, 2007
Table DQ.10: Distribution of women by time since last birth
Distribution of women aged 15-49 years with at least one live birth (weighted), by months since
last birth, Nigeria, 2007
Quality assessment study of the data has confirmed a number of quality problems in MICS Nigeria 2007. In the
following paragraphs we set out these problems offering the likely causes as well as some of the possible
implications for data quality and accuracy of estimates of characteristics and indicators emanating from the data
Age Heaping
Large amount of heaping exists at ages with digits ending in 0 and 5 except at age 15.This exception is not genuine
being yet evidence of some other quality problem (Table DQ.1 Table DQ.5 and, Figure DQ.1)). Illiteracy
particularly un respect of women respondents, cultural bias for figures ending with 0 and 5, cultural practice that
counts in 5s, poor book keeping habit, burden of length of questionnaire, and other reasons Age heaping is also
evident in the male age data. This problem could lead to a false impression of the age structure resulting from some
over-representation of persons of ages ending in digits 0 and 5. There could be bias in weighted estimate of any
characteristic that depends on age structure e.g. mortality rate. Effect is less in respect of characteristics that depend
on age grouping where the ages ending 0 or 5 are less important and where differentials in respect of the
characteristics of interest about the heaps are trivial.
Out-Transfer of Ages of Women and Children
Large out-transfer of children from target group 0-4 year old (Table DQ.3, Figure DQ.2) and of women from the
target group 15-49 year-old was evident; a proof is the unlikely pyramidal structure of age distribution; some
children of genuine age 4 (or even lower) must have had their ages recorded as 5 or more years; also a good number
of women with true age 15 years or higher must have been recorded as 14 years old or younger; and some women
truly aged 49 years or lower have had their ages recorded as 50 or higher (Figure DQ.3). Possible effects of the outtransfers
could include a substantial detraction from the quality of the data and from the general accuracy of those
indicators that use differential weights that are derived from the relative frequency distribution of the ages. This
means that children aged 4 years and women aged 15 and 49 years respectively may have been poorly reflected in
the sample; it means that these children and women have been under-sampled, that is children aged 0-4 and women
aged 15-24, 45-49 and 15-49 may have been quite severely under-represented.
Estimates of group characteristics of the children under 5 and of women in each of the affected age groups stand
adequate and credible as long as sample size posed no serious precision problem. But combined estimates derived
from weighted estimation would have problem of bias particularly if there are differences across ages and age
groups.
Lower Response Rates Among Younger Women.
Differential response rates are noted across age group, lower among the younger women aged 15-24 years (Table
DQ.2) (Figure DQ.4); this translates in to differential representation and data accuracy across the age groups. The
likely effect includes a distortion of the weights and a bias in estimates. But response rate ranged from 86 to 95
percent; bottom 86 percent seems quite adequate though quite less than MICS3 suggested bottom figure of 90
percent The fear is that some bias in favour of the older women may result particularly in combined estimates across
ages; inevitably, this could detract from the accuracy of results particularly if the non-respondents coincide with a
sub-group with characteristics that are distinct from the rest of the population.
Incomplete information on dates, month, year of birth and marriage
Age data featured disproportionately large amount of ‘missing’ and ‘don’t know’ in data on dates of marriages of
women and births of children and adults. This is a a problem of the poor or the uneducated or the rural person the
poor; it is a problem aggravated by characteristic inadequate birth registration and poor record keeping habits. The
cost could be a substantial reduction in effective sample size impacting adversely on the accuracy of estimates of
child outcomes that require an accurate recollection of dates of birth of the child and of landmarks in child history
e.g. weaning, breastfeeding food supplementation, vaccination, pre-school development. Good recollection of dates
of events is also a vital requirement for quality of results on mortality rates.
Large Over-Age Children in Pre-School and Primary Schools
There are large numbers of household members’ age 8+ attending pre-school, similar unexpected numbers of
household members at quite unexpected ages are attending other levels of schools including the primary . If these
are confirmed as errors, then they probably suggest incorrect trend and a misrepresentation of pre-school
development and primary school attendance; it means an under-estimation of primary school attendance ratio and a
general loss of accuracy in the results
On the other hand, it is evident that there is a strong diagonal feature if we take the ages in groups e.g. 5-7, 6-8, etc.,
this suggests there could be some late starts in primary school enrolment, a feature that splits over into the higher
grades of the primary school and beyond.
Large Male-Female Ratio
Sex ratios at birth are consistently above the expected 1.05-1.06 level (Tables DQ.1 & DQ.9 and Figures DQ.4-
DQ.5) This usually indicates that some female children are not declared. This criticism suggests possible undersampling
of the female and in its wake an under-representation of the female children; it would also suggest a tilt to
male sex domination beyond the norm.
Under-declaration of female children necessarily distorts sex ratio figure and gender balance; an under-sampling of
the girl-children reduces the sample size and the precision of estimate of girl-child outcomes. It could also affect
estimates of sex differentials.
Large Exclusion o Children in the Calculation of Anthropometrical Child Outcomes
A large number of children are excluded from the tabulations on malnourishment, because of missing data (Table
DQ.5) Some 29 percent of all children under 5 are excluded from the analysis. This figure includes 11 percent who
were excluded because the weight and/or height measurements were out of range, and 17 percent for who date of
birth was incomplete; the exclusions were 17% due to missing date or year of birth and other causes. The missing
cases could as well be children of the most poorly educated mothers or children in the poorest wealth index
quintiles. Hence malnutrition could be more prevalent and more intense among them. In effect, the true state of
malnutrition in the country could be more serious than depicted by the data
Heaping of height and weight measurements
Considerable heaping of height and weight measurements around decimal point 0 and 0.5 most especially around 0
has been observed. Apparently figures ending 0.1, 0.2, 0.3, 0.4 were rounded down to next whole number below.
Figures ending 0.6, 0.7, 0.8, 0.8 were rounded
up to the whole number above while figures ending 0.5 were left alone because canvassers would not know whether
to round up or down (Figures 8a 8b). The errors here could mutually cancel out; the mean and the standard deviation
may not be significantly distorted, and the bias minimal. But if the individual measurement is considered against an
interval to decide the level of malnourishment of the individual child, then the effect of the difference of
magnitude 0.1 to 0.4 arising from rounding up or down of the individual measurement may be more than trivial
The extent of distortions associated with the tabulated results would depend on the extent to which differences of 0.1
to 0.4 in measurements of individual weight and height respectively influence the placement of an individual on the
weight for age (underweight), height for age (stunting) and weight for height (wasting) scales respectively. Weights
are measured in kg and height in cm; it is unlikely that differences of magnitude 0.1 – 0.4 cm in height and 0.1-0.4
kg in weight would make any significant difference in these placements.
Low Child Mortality Rates
Estimates of infant and under-5 mortality rates by MICS Nigeria 2007 are low.. Some inconsistency,
incomparability and incompatibility with previous survey results is suspected. Criticism that the figures are underestimates,
if well-founded means that child deaths have been under reported, or age structures of the children and of
the ,others have been misreported or that the calculating method is sensitive to such misreporting.