Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Appendix: Variable description and composite index construction


The variables presented here were based on a subset of data from the Ford Foundation-funded intervention project, Changing Caretaking Behaviors to Improve Child Growth and Development (Satoto and Colletta 1987). The project consisted of three phases: (1) sample selection (including both intervention and control groups) and collection of baseline information in 1987; (2) 18 months of in-home intervention with the kaders and mothers, and (3) follow-up data collection and analysis in August-November 1989. The data used in this study were based on the follow-up data. A total of 304 children over two years of age were selected for this study (Chomitz 1992).

Social resources

MOTHER'S YEARS OF EDUCATION (MOMED). This variable referred to the total years of mother's education (value ranged from 0 to 16). A summary of descriptive statistics is as follows:

No education (0 years) 1.0%
Primary school (2-6 years) 73.1%
Secondary school (7-9 years) 12.9%
High school (10-12 years) 12.3%
College (13-16 years) 1.7%

MOTHER'S LITERACY (LITERACY). This variable was composed of two survey questions (coded from 1 to 4):

1. Reading abilities of mothers (MOMREAD)

Very fluent 0.8%
Fluent 16.2%
Slow 61.8%
Cannot read 21.2%

2. Writing abilities of mothers (MOMWRITE)

Very fluent 22.5%
Fluent 54.3 %
Slow 22.5%
Cannot write 0.7%


MOTHER'S SOCIAL NETWORK (SOCNET). This variable was composed of two survey questions:

1. Number of organizations mother joins (MOMNOORG) ranged from 0 (none) to 5

2. How frequently mother meets friends (ISOLATE) ranged from 1 to 4:

Never 1.2%
2-4 times monthly 6.2%
2-3 times weekly 18.2%
Every day 74.4%

The definition of SOCNET is: SOCNET = SUM (MOMNOORG, ISOLATE).

Material resources

QUALITY OF HOUSE (HOUSING). This variable was composed of three survey questions (value ranged from 1 to 4):

1. Building material of front of house (HOUSE1)

Not brick/stone 49.8%
Half brick 17.4%
Brick without windows 9.3%
Brick with windows 23.5%

2. Floor of living room (HOUSE2)

Earth 71.1%
Half earth/half cement 7.4%
All cement 4.7%
Tiles 14.1%

3. Lighting in living room (HOUSE3)

Nothing, or sentir 2.4%
Kerosene lamp 30.3%
Petro Max/City electricity 59.8%
Diesel: own generator 7.5%

The definition of HOUSING is: HOUSING = SUM (HOUSE1, HOUSE2, HOUSE3).

WEALTH POSSESSIONS (TOTITEM). This variable was composed of household possessions (nine items). The nine items were added up to form a single variable (TOTITEM) weighted by factor scores based on ranks of status of each item (Chomitz 1992) (0 = no, 1 = yes). The percentages of "yes" scores for each item were as follows:

1. Car ownership 0.8%
2. Motorcycle ownership 21.4%
3. Sewing machine ownership 12.8%
4. TV ownership 30.0%
5. Sofa ownership 17.2%
6. Radio-cassette ownership 65.7%
7. Bicycle ownership 56.0%
8. Petro Max ownership 58.1 %
9. Mattress ownership 95.7%

The definition of TOTITEM is: TOTITEM= (ITEM1*3) + (ITEM2*3) + (ITEM3*3) + (ITEM4*3) + (ITEM5*2) + (ITEM6*2) + (ITEM7*2) + (ITEM8) + (ITEM9*2).

Health practices

TOTAL IMMUNIZATION (TOTIMM). TOTIMM refers to total immunization received by the child during his/her entire life. It is composed of three immunization types: diphtheria/pertussis/tetanus (DPT), polio, and bacillus Calmette-Guérin (BCG). DPT and polio have three boosters from which the total shots were added into TOTDPT and TOTPOLIO. TOTDPT and TOTPOLIO have values ranging between 0 and 3 (0 means no immunization at all and 3 means a complete course was received). For BCG, 0 means no immunization and 1 means immunization received. Scores were as follows:

0: 8.1%
1: 15.4%
2: 10.4%
3: 56.3%
0: 24.7%
1: 5.8%
2: 6.9%
3: 62.5%
0: 6.3%
1: 93.7%

The composite index TOTIMM is defined as follows: TOTIMM = SUM (TOTDPT, TOTPOLIO, BCG).

VITAMIN A SUPPLEMENTATION (VITASUP). VITASUP is defined as the total number of vitamin A tablets received by the child over his/her entire life. The values ranged between 0 (none) and 6 (six tablets received). The percentages of children scoring 0-6 were as follows: 0, 7.3%; 1, 2.7%; 2, 10.8%; 3, 14.3%; 4, 13.1%; 5, 10.8%, and 6, 40.9%.

Community endowment

HEALTH INDICATOR (HLTHIND). The HLTHIND variable consisted of community-level indicators on crude birth rate (CBR) and under-five mortality rate (U5MR). These variables were expected to reflect the quality of programme or facility available in the community. The CBR (per 1,000 population per year) was 39 (35.7%), 67 (56.1%), and 120 (8.2%). CBR was transformed into 0 (if above the mean) and 1 (if below the mean).

The U5MR (per 1,000 population) was 3 (35.7%) and 4 (56.9%). U5MR was recorded as 0 and 1 (0 = 4; 1 = 3).

The HLTHIND variable was defined as follows: HLTHIND= SUM (CBR, U5MR

HEALTH FACILITIES AVAILABLE IN THE COMMUNITY (FACILITY). FACILITY IS composed of three information variables: the number of traditional birth attendants (TBA), nutrition groups (NUTGRP), and health facilities (HLTHFAC) available at the community level. The descriptive statistics of these variables are presented below. Since each variable has a different scale measurement, they were transformed into 0 and 1 values (1 = higher than average; 0 = lower than average).

The TBA scores (per 1,000 population) were 6.6 (56.1%), 18.30 (35.7%), and 27.80 (8.2%).

The NUTGRP scores (per 1,000 population) were 6.3 (56.1%), 9.50 (35.7%), and 10.40 (8.2%).

The HLTHFAC scores (per 1,000 population) were 2.2 (49.7%), 4.1 (35.7%), and 6.0 (7.2%).

The FACILITY variable was defined as follows (after all variables were transformed): FACILITY= SUM (TBA, NUTGRP, HLTHFAC).

Feeding practices

PROTEIN ADEQUACY (PROTEIN). Protein adequacy is also a measure of energy adequacy, which is why we call this term ProtCal. We used protein as a proxy for protein and energy because it was more highly correlated with child developmental outcomes than was caloric adequacy. Protein adequacy was calculated by Chomitz (1992) based on the total portions of foods per day collected by the food frequency questionnaire (FFQ) developed by Chomitz (1992). The protein intake was adjusted downward to correct for the digestibility of the diet and quality of protein consumed, in order to represent the estimated biological value of the diet. A Filipino mixed diet (digestibility relative to reference proteins = 93) was chosen as the most similar diet for which information was available (for detailed description see Chomitz 1992, 113). The percentage recommended dietary intake ( % RDI) was calculated, in order to present the data in a more readily interpretable manner and to control analytic variables for age, weight, and characteristics of the diet.

The protein intake value collected by the FFQ was compared with estimated requirements for actual weight and US National Center for Health Statistics (NCHS) weight for age and sex. Thus, PROTEIN was defined as follows: ProtCal= (Available Protein/RDI of Protein)*100. The mean and the median of this variable were 255 and 234%, respectively.

VITAMIN A ADEQUACY (VITA). Total vitamin A intake was calculated by Chomitz (1992) based on 24-hour recall information. The RDI of vitamin A was based on the United States Recommended Dietary Allowance (National Research Council 1989) estimate (RE/day) according to age categories. The %RDI of vitamin A (VITA) was estimated as follows: VITA = (available vitamin A RE/RDI of Vitamin A)*100. The mean and the median of this variable were 269 and 187%, respectively.


MATERNAL AFFECTION (AFFECT). Factor analysis was used as the basis of creating the AFFECT scale using the US-based Caldwell HOME inventory of mother-and-child interaction. Two factors related to warmth and affection emerged from the factor analysis (WARMTH and ACCEPT). Since these factors were highly correlated, they were combined into a single factor called AFFECT.

The percentages of positive responses for the WARMTH and ACCEPT factors were as follows:

Home 26. Parent holds child close for 10-15 minutes per day 70%
Home 30. Parent praises child's qualities twice during visit 49%
Home 31. Parent caresses, kisses, or cuddles child 49%
Home 40. Parent introduces visitors to child 73%
Home 54. Parent does not slap or spank child 95%
Home 55. No physical punishment used during past week 88 %.

The AFFECT variable is defined as follows: AFFECT = SUM (WARMTH, ACCEPT).

MOTHER'S TIME SPENT WITH CHILD (HRSWKID). The total amount of time the mother spent with the child per day (HRSWKID) consisted of five items: (1) mother out of home with the child (HRS1); (2) mother doing housework with the child (HRS2); (3) mother sleeps with the child (HRS3); (4) mother feeds the child (HRS4); (5) mother plays with the child (HRS5).

The variable HRSWKID is defined as: HRSWKID = SUM (HRS1, HRS2, HRS3, HRS4, HRS5).

The mean and median hours per day the mother spent with the child were 14.6 and 15, respectively.

Academic stimulation

ACADEMIC STIMULATION AND PHYSICAL ENVIRONMENT (STIMUL). Factor analysis was used to create a STIMUL scale using the USbased Caldwell HOME inventory of mother-and-child interaction. Two factors related to learning and teaching environment emerged from the factor analysis (EDUPLAY and VARIETY). Since these factors were highly correlated, they were combined into a single factor called TEACH.

The percentages of positive responses for the EDUPLAY and VARIETY factors were as follows:

Home 3. Child owns five tapes 16%
Home 8. Books visible in home 65%
Home 7. Child owns books 63%
Home 10. Magazine subscription 7%
Home 43. Child has real or toy musical instruments 22%
Home 45. Child has been on trip more than 50 miles in past year 34%
Home 46. Child has been to museum during past year 9%
Home 49. Child's art work 9%.

The STIMUL variable is defined as follows: STIMUL= SUM (EDUPLAY, VARIETY).

LEARNING AND TEACHING ENVIRONMENT (TEACH). The same procedure as before was used to construct the TEACH variable. Based on factor analysis loading factors using Caldwell's HOME inventory for measurement of the environment for preschool children, three related factors emerged, namely LEARN, LANGUAGE, and DISCIP. Some items in LANGUAGE are not culturally appropriate (see section on Javanese family literature review): for example, a child is not supposed to talk to parents in the presence of visitors (e.g. interviewers); therefore, Home 27 (parent converses with child twice during visit), Home 28 (parent answers child's questions verbally), and Home 29 (parent responds verbally to child's speech) were eliminated. This elimination improved the correlation of LANGUAGE variables with some child developmental outcome.

The percentages of positive responses for the LEARN, LANGUAGE, and DISCIP factors were as follows:

Home 13. Encouraged to learn alphabet 78%
Home 37. Encouraged to learn to read a few words 72%
Home 14. Parent teaches verbal manners 97%
Home 38. Child must wait for mealtime 22%
Home 39. TV is used judiciously 45%.

These three variables were conceptually related and were highly correlated. Therefore, they were combined into a single variable called TEACH: TEACH = SUM (LEARN, LANGUAGE, DISCIP).


EMOTIONAL SUPPORT RECEIVED BY THE MOTHER (EMOSUP). This variable was the sum of the following responses. Each response was coded between 0 and 2 (0 = not happy; 1 = happy; 2 = very happy):

  • 1. Is mother satisfied with family help caring for the house?
    2. Is mother satisfied with family help in child rearing?
    3. Is mother satisfied with moral support she received from others?
    4. Is mother satisfied with frequency of meeting with relatives and friends?
  • FATHER'S INVOLVEMENT IN CHILD CARE (FATHER). This variable was made by using the information on substitute caretakers. If the father was the first substitute caretaker, a value of 3 was assigned; if he was the second substitute caretaker, a value of 2 was assigned; if he was the third substitute caretaker, a value of 1 was assigned; if the father was none of those, a value of 0 was assigned. A percentage of 49.4% of the Javanese fathers were involved in child-care activities. FATHER: no involvement, 50.6%; SUB1, 17.4%; SUB2, 14.3%; SUB3, 17.8%.


    A child's growth was determined from repeated measures of weight (kilograms) and height (centimetres), converted to ratios of weight-for-age (WAZ) and height-for-age (HAZ) and compared with the US NCHS reference population data according to procedures recommended by WHO. The means of WAZ and HAZ were -1.53 and -2.24, respectively.

    Mental development

    Stanford Binet IQ scores (IQ) and Vineland's Social Quotient scores (SQ) were computed. The means of these scores were 89.1 and 102.7, respectively.


    The variables presented here were based on a subset of data from the Positive Deviance in Nutrition Research Project funded by UNICEF, New York, conducted in Lagos, Nigeria (Aina et al. 1992). This project focused primarily on urban and semi-rural children in Lagos State, south-western Nigeria, with a small rural subsample. A total of 211 two-year-old children were chosen.

    Social resources variables

    MOTHER'S LITERACY (LITERACY). Literacy refers to the mother's last year of school completed (LSTYR), and the literacy of the mother (reads Yoruba and reads English). The definition of LITERACY is LITERACY = LSTYR + RDENGLSH + RDYORBA. The results were as follows:

    1. LSTYR (coded from 0 to 5)
    No formal education (35%)
    Some primary (17 % )
    Primary completed (27%)
    Some secondary (11%)
    Secondary completed (11%)
    2. Literacy of mother (1 = yes) (0 = no)
    Reads Yoruba (56%)
    Reads English (37%).

    MOTHER'S EXPOSURE TO MEDIA (MEDIA). Mother's media exposure consisted of three questions:

    1. Number of states in Nigeria correctly identified (STAT), coded from 0 to 2:

    None 4%
    One to three 22%
    Four or more 74%

    2. Names of radio programmes identified (RADPGMM), coded from 0 to 3:

    None 36%
    One to two 38 %
    Three 21%
    Four or more 6%

    3. Numbers of TV programmes identified (TVPGM), coded from 0 to 3:

    None 13 %
    One 13%
    Two to three 45%
    Four or more 30%.

    The composite variable MEDIA was computed as follows: MEDIA = STAT + RADPGM + TVPGM.

    LOCATION (LOCATN). The location variable consisted of three categories (coded from 0 to 2):

    1. Rural 11%
    2. Semi-rural 41%
    3. Urban 48%.

    Material resources

    Two variables, HOUSING and LFDBUDGE, were chosen to mea sure the latent variable Material Resources. These two variables were significantly correlated.

    HOUSING. The quality of house composite index consisted of the following information:

    1. Housing type (TIPE) coded from 1 to 3

    Face-to-face 74%
    Compound or flat 9%
    Detached 17%

    2. Wall construction (WALLS) coded from 1 to 3

    Mud 13 %
    Corrugated iron 9%
    Cement blocks 79%

    3. Roof construction (ROOF) coded from 1 to 4

    Thatched 2%
    Asbestos sheets 32%
    Corrugated iron 65%
    Concrete 1%

    4. Floor construction (FLOOR) (1-4)

    Mud 2 %
    Wood 2%
    Cement 80%
    Linoleum 15%
    Tiles 1%

    5. Number of rooms (RMS) (1-3)

    One 71%
    Two 19%
    Three to four 10%

    6. Housing condition (CONHOU) (1-4)

    In very bad shape 13%
    In need of repairs 51%
    In good repair 33%
    Paint, furniture looks new 3%

    7. Electricity in house (ELECTR) (1-2)

    No 25 %
    Yes 75%.


    FOOD BUDGET (LFDBUDGE). Food budget consisted of the mother's contribution for food (CASH1), the father's contribution for food (CASH2), and the total amount all others contributed for food (CASH3): FDBUDGE = sum (CASH1, CASH2, CASH3); LFDBUDGE = In(LFBUDGE).

    Feeding practices

    FOOD BELIEF (FDBELIEF). Meat was chosen in order to investigate the mother's perception of the proportional reward function. Meat was selected because it was the food most frequently and emotionally referred to when parents articulated the traditional value system with regard to children, food, and spoiling.

    The question investigating belief towards meat was, "Is there any reason why you don't think a child of this age should have more meat?" The percentages of positive responses (coded as 0) were as follows (a code of 1 means negative response):

    1. Spoil child's moral character 66%
    2. Cause a child to steal 31%
    3. Can spoil the child 46%.

    The composite index of FDBELIEF is defined as follows: FDBELIEF = SUM (1,2,3)

    FOOD INVESTMENT (FDINVEST). The food investment variable consisted of two questions:

    1. Imagine that you have plenty of money for food. How big a piece of meat do you think is the right amount for a two-year-old child for one meal?

    2. Imagine this is a piece of meat enough for one meal, and the child's father were eating with you. Please show how much meat each person in the family would receive. Be sure to indicate how much the two-year-old child will get compared to the other children.

    An 11 x 9-inch board, one-third of an inch thick and varnished a beefsteak brown colour was developed to represent a slab of meat. This meat-board was deliberately made rather large in size to convey the notion that plenty of meat was available. It was marked by a grid of lines forming 396 half-inch squares, which represented pieces of meat. Mothers were asked to indicate their answers to questions about meat distribution by drawing on this board with chalk.

    These two components were combined into the right amount of meat for a two-year-old. The composite variable is right amount of meat for two-yearold times sums of weekly frequencies of snacks, fruits, and other animal foods. The formula is: sum of the weekly frequency of intake of animal food, snack, and fruit, weighted by the portion size of meat appropriate for a twoyear-old child. Sum (TOTANI, TSNACK, NFRUITOT) * (MTWOYRS). This variable is transformed into a natural logarithmic transformation.

    Hygiene practices

    The cleanliness of the environment consisted of three indicators Cleanliness, Hygienic information, and Water source.


    1. Debris on floor, inside or out.
    2. Animal faeces (ANIMFC) (no = 1) (yes = 0)

    No 87%
    Yes 13%

    3. Human faeces (HUMFC)

    No 99%
    Yes 1%

    4. Spoiled food (SPLFD)

    No 94%
    Yes 6%

    5. Dirt or other (DIRT)

    No 33%
    Yes 67%.


    HYGIENIC INFORMATION (SANIT). Hygienic information refers to the following (yes = 1) (no = 0):

    1. Was bowl cleaned before using (BWLCL)?

    Yes 98%
    No 2%

    2. Were the utensils clean (UTENSCL)?

    Yes 97%
    No 3%

    3. Did mother clean hands (HNDCLN)?

    Yes 90%
    No 10%.

    4. Did mother clean her hands (MOMHND)?

    Yes 81%
    No 19%.


    WATER SOURCE (WATERSC). The water source variable has five categories ranging from 1 to 5:

    1. Stream 9%
    2. Well 9%
    3. Rain or tank 6%
    4. Borehole 12%
    5. Tap 64%.


    MATERNAL AFFECTION (WARM). This variable (WARM) consisted of a sum of interviewer rating of maternal behaviour and affection towards her child during the Bayley test situation. WARM is the sum of eight variables (each variable was coded from lowest to highest, or 15). The percentages of positive responses for each category are shown in table 8.1A.

    MATERNAL VERBAL RESPONSIVENESS (VERBAL). This variable (VERBAL) is a sum of observed maternal verbalizations with her child. Each was coded from 0 to 2.

    1. Simple words that child could repeat

    Never 53%
    Sometimes 35 %
    Usually 10%

    Table 8.1A Percentage of mothers scoring one (lowest) to five (highest) on eight variables of Maternal Affection (WARM)

      Percentage of mothers with score
    Variable Lowest Low Average High Highest
    Comfort 2 14 29 46 8
    Encouragement 2 12 26 55 3
    Eagerness 1 7 25 57 9
    Emotional 1 7 17 64 10
    Visible signs 1 13 19 66 10
    Responsiveness 1 6 28 61 5
    Response to mother 2 16 20 53 8

    2. Were sounds without meaning

    Never 89%
    Sometimes 7%
    Usually 4%.

    CHILD-CARE ARRANGEMENTS (CHCARE). This variable included the quality of primary caregivers, secondary caregivers, total number of regular caregivers, and mother's proximity to the child when she worked. They were ranked using criteria consistent with the Caldwell HOME inventory. The Zscores with these separate indicators were summed. In this sum, full weight was given to primary care and number of caregivers and half to secondary care and proximity to the mother.

    Academic stimulation

    TEACH CHILD (TEACH). The TEACH variable consisted of seven variables that related to the following questions (each was coded 1 = yes; 0 = no):

    1. Do you teach your child the ABC? (ABC) (69% yes);
    2. Does anyone in the family ever read him/her stories? (RYES) (25 % yes);
    3. Does anyone read stories? (STY) (38% yes);
    4. Mother's play with child - social play (SOCT) (25% yes);
    5. Mother's play with child - technical play (TPLA) (9% yes);
    6. Do you teach child to put away toys? (90% yes);
    7. Do you teach child to wash hands? (90% yes).

    TEACH is the sum of these variables.

    AVAILABILITY OF TOYS (TOYS). This variable is the sum of variables that relate to the following questions:

    1. Does your child have any of the following toys? (yes = 1; no = 0):

    - Football, plastic ball, toys with wheel, rocking horse, toy airplane, wind-up toy, plastic rattle, plastic dolls, plastic building toys, puzzles, crayons, jewellery (not including earrings), stuffed animal, other toys

    2. What other things does he/she like to play with? (yes = 1; no = 0):

    - Bottle covers, special stones (sticks), empty tins, plastic containers, school box, papers or newspaper, plastic bag, special cloth, sand, and water.

    Modernizing lifestyle

    MODERN BEHAVIOURS (MODBEHAV). This variable was the sum of variables relating to the following information (coded as 1 = yes and 0 = no). These variables loaded onto a single factor in factor analysis:

    1. Eats with mum and dad at least twice a week (78% yes);

    2. Caresses the child (48% sometimes = 0, 52% usually = 1);

    3. Uses modern medical treatment (35% yes);

    4. No safety hazard (28% yes);

    5. No physical punishment, shouting, or threats observed toward the child during three-hour home visit (55% yes);

    6. No housework required from two-year-old (83% yes).

    FATHER LIVES WITH MOTHER (PALIVE). PALIVE consisted of four categories (coded from 1 to 4):

    Father lives with mother:

    1. Rarely or never 10%
    2. Less than half-time 11%
    3. More than half-time 5%
    4. All the time 74%.

    Child's personality

    ELICITING BABY (ELICBABY). ELICBABY consisted of observer ratings of infant behaviour and affect during the Bayley test. ELICBABY is the mean of five components. Each component was coded from 1 to 5 (lowest to highest) (table 8.2A).

    THE EASINESS OF THE CHILD HANDLED BY THE INTERVIEWERS (COOPBAB). COOPBAB is the reciprocal of the mean of several components (ranged from 1 to 5 or lowest to highest) (table 8.3A).


    ANTHROPOMETRIC INDICATORS (WAZ AND HAZ). Weight, height, and age data were collected as recommended by WHO. Birth dates were determined from birth certificates. Weights were taken to the nearest 0.1 kg using a hanging scale with a capacity of 25 kg. A wooden stadiometer built by the project was used to measure height in cm (2-6 years).

    Raw weights and heights were converted to Z-scores (WAZ and HAZ) of the NCHS-WHO standards using software provided by the United States Centers for Disease Control. Two centimetres were added to the standing heights of the children just below two years, whose measurements were treated by the CDC program as recumbent lengths. The means of WAZ and HAZ were-1.78 and-2.29, respectively.

    Table 8.2A Percentage of babies scoring one (lowest) to five (highest) on five components of infant behaviour and affect (ELICBABY)

      Percentage of babies with score
    Component Lowest Low Average High Highest
    Eagerness 5 19 24 42 11
    Dull to gleam 1 23 27 40 9
    Attractiveness 4 14 32 43 8
    Attraction to child 6 17 40 31 6

    Table 8.3A Percentage of babies scoring one (lowest) to five (highest) on five components of ease of handling (COOPBAB)

      Percentage of babies with score
    Component Lowest Low Average High Highest
    Angry 31 33 12 18 5
    Irritable 20 31 24 20 6
    Affective 18 36 18 23 5
    Task orientated 13 37 16 28 6
    Provocative 20 37 20 20 3

    Mental development

    The Bayley mental development index (MDI) and Bayley physical development index (PDI) were used to construct these variables. The means of MDI and PDI were 92.3 and 103.6, respectively.

    Contents - Previous - Next