Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu



Variability of total energy and protein intake in rural Bangladesh: Implications for epidemiological studies of diet in developing countries


 

Alberto Torres, Walter Willett, John Orav, Lincoln Chen, and Emdadul Huq

 

Abstract

It has been recognized for some time that nutrient intakes show large day-to-day variations within the same individual in industrialized countries. However, little attention has been given to the quantification of within-and between-person components of the variation in developing countries. Within-person variability determines the number of measurements of dietary intake per person that will be needed to adequately represent the dietary intake of an individual over a period of time. A large within-person variation relative to the variation between persons in the intake of a nutrient decreases the strength of the observed association of the latter with an outcome of interest.

This study quantified components of variation for 834 people of both sexes and various age groups and found that, although the diet in rural Bangladesh includes a limited number of food items-over 80% of calorie and protein intakes are provided by cereals, with rice being the main staple-the daily variability of nutrient intake in each individual over a year was as large as that reported for industrialized countries. The authors estimate that about seven days would be needed to assess total energy and protein intakes adequately, while much longer measurement periods would be needed for animal protein.

 

Introduction

The implications of variations within the individual for physiological measurements have been extensively described and analysed for diet, blood pressure, and blood lipids [1-9]. It has been recognized for some time that dietary intake varies greatly from day to day in the same individual [10-12]. Some relatively recent studies have quantified components of variation for dietary intake, but these have been limited to industralized countries [13-17].

The authors are affiliated with the Harvard University School of Public Health in Boston, Massachusetts, USA.

All of them show that estimation of a person's true dietary intake over a period of time requires repeated measurements. A practical consequence of this is that isolated measurements are poor estimators of long-term dietary intake. The error that results from taking isolated measurements cannot be diminished by increasing sample size but only by increasing the number of measurements for each individual [15].

Little attention has been given to the quantification of within- and between-person components of the variation in developing countries. There are several reasons why it is important. First, if within-person variability is small, a few careful measurements per individual will suffice to represent true dietary intake over a period of time; but, if it is large, repeated measurements across time will be required in order to estimate intake with the same level of accuracy. Therefore, knowledge of the within-person component of variation will allow the choice of appropriate study designs and strategies for measuring dietary intake, as well as for evaluating the meaning of dietary information already collected. Second, the quantification of the contribution of other factors (e.g. seasonality) to the variation is important to determine the need for repeated measurements across seasons. Third, the effect of large variation within persons on the measurement of the association between dietary intake and an outcome measure is to decrease the strength of the association [9; 18], and, since epidemiological studies in developing countries often estimate dietary intake using a single 24-hour recall questionnaire, it is important to test the implicit assumption they make that the variation within persons is low.

Finally, dietary intake in rural areas of developing countries is both qualitatively and quantitatively different from that in industrialized countries. It could be expected that within-person variation in diet might have a culturally specific value [15]. Surveys have shown that more than 80% of calorie and protein in take in rural Bangladesh is provided by cereals [19]. Rice is the main staple, although wheat and millet may be important for poor families during periods of food scarcity. Although the diet includes a limited number of food items, the daily variation of nutrient intake in each individual over a period of time may not be small.

Several factors may affect the variability of within-person nutrient intake. Seasonal changes are potentially important in developing countries [20], including rural areas of Bangladesh [19; 21-24]. The Matlab area in southern Bangladesh has three seasons- monsoon (June-October), cool winter (November-February), and the hot, dry season (March-May)- each accompanied by specific cropping practices [24]. In industrialized countries, although there are seasonal components to the consumption of some micronutrients, seasonality plays a minor role in the variation of total energy intake [25; 26] and probably of most specific nutrients. Within-person variability that is not explained by any other factor may also be an important component of the total variation.

One difficulty in assessing the components of variation is the usual lack of repeated measurements of diet for the same individuals. The present study examines repeated measurements of dietary intake for more than 800 persons over one year. The number of individuals studied makes it possible to analyse for specific age groups with a level of detail that has not been previously available.

The aims of this study were to estimate within- and between-person components of the variation of total energy and protein intake in a rural population of Bangladesh, to test the importance of seasonality for diet, and to derive conclusions about study design and the interpretation of dietary data.

 

Methods

The data

Source: The Matlab study

The data for the present analysis come from a field study conducted between June 1977 and August 1978 in Matlab Thana, the experimental field station of the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), the aim of which was to elucidate determinants and consequences of physical growth of children under five years of age.

The 135 households that were included were non-randomly selected from four villages to reflect a wide range of socio-economic status in the study population. In total, 954 subjects were included: 207 children and 747 other household members (table 1). The study included 44 families owning less than 0.5 acres of land, 51 owning between 0.5 and 2.0 acres, and 40 owning more than 2.0 acres.

TABLE 1. Weights and heights of the study population by sex and age group at the beginning of the study

Sex and

age group

(years)

N Weight (kg) Height (cm)
Mean SD Mean SD
Male 486a
0 - 4 113a 9.1 2.2 80.0 9.8
5-14 132 19.0 5.0 116.2 12.5
15-44 181 45.3 7.6 158.7 8.3
345 60 45.3 6.0 160.8 5.9
Female 468a
0-4 94a 8.7 2.5 79.0 11.0
5-14 125 19.2 6.2 116.0 14.1
15-44 194 40.8 5.0 148.2 5.2
45 55 37.5 9.8 145.4 9.1

a. These figures include 66 male and 54 female breast-feeding children who were excluded from the analysis.

Data collection

One 24-hour dietary-intake measurement was obtained for all persons in each household every two months, with a total of six measurements for each person. The survey team consisted of two female dietary workers with high school education, three female field assistants, and one male helper. On each visit day, the selected house was visited by one dietary worker and one field worker.

Food intake was measured by 24-hour weighing of food, both before and after cooking, in addition to volumetric measurement. All large cooking vessels and serving utensils in the households were measured at the beginning of the study using standard chemistry laboratory Teflon volumetric cylinders (2 litres, 1 litre, 500 ml, 250 ml, 50 ml), and the volumes were recorded. On the day of the home visit, the workers arrived before the first meal was prepared. After standardizing and calibrating their scales (Salter, 5-kg range, discrimination 2 g), they weighed the leftover food from the previous day. All foods to be consumed were weighed and measured before and after cooking.

Individual consumption was estimated by recording the number of portions of each food item served to each family member. A homogeneous mixture of dry ingredients in the cooked food was assumed to estimate actual consumption. Individual plate waste and leftovers were measured. Meals consumed outside the household and snacks were assessed by recall conducted twice daily. Then total 24-hour consumption was estimated.

The sum of individual intakes was recorded as the household intake level. The intake of macronutrients was computed using standard food conversion tables from India [27].

Description of the data

Six 24-hour measurements of total energy and protein intake for each of 954 subjects were available for analysis. One hundred twenty breast-feeding children were excluded from the analysis because of lack of information on breast-milk consumption. This reduced the study population to 834 subjects. Approximately 10% of the dietary measurements were missing at each visit; all those available were included in the analysis, however.

Information on protein intake was classified for total protein, protein of cereal origin, and protein of non-cereal origin, mostly animal protein. The subjects were classified by sex and four age groups: (0-4, 5-14,15-44, and 45 years and over). The mean intake for each age and sex group at each of the six measurements is presented in table 2. The figures are higher than those estimated by previous surveys in other areas of Bangladesh [19].

 

Methods of analysis

The goals of the analysis were to quantify different components of the variation of protein and total energy intake, specifically, the within- and between-person components, as well as seasonality. Each bimonthly dietary measurement was regarded as an estimate of the individual's true intake during the year, subject to random variation due to within-person variation, with additional allowance for seasonal effects in some models.

The mixed analysis of variance (ANOVA) model used in the analyses can be summarized as where Yij represents the dietary measurement on visit j day for participant i, u is the average population intake, subject) is the random-effect term that reflects between-person variability and has a distribution with mean 0 and variance, factor Xij is a main-effect term used in some models for seasonality which identifies the fixed effect that measurement visit j has on the diet of subject i, and Eij is the unexplained random within-person variability, having a distribution with mean 0 and variance .

Yij = u + subjecti + factor Xij + Eij

TABLE 2. Energy and protein intake by age group and sex

Age group (years), sex, and visit  

Energy (kcal)

Protein (g)

Total

Cereal

Non-cereal

N Mean SD Mean SD Mean SD Mean SD
0-4                  
male 47                
1   1,138 395 32 16 21 8 11 13
2   1,077 359 35 15 21 8 14 12
3   1,037 346 27 10 20 7 8 5
4   1,025 423 24 10 17 7 7 5
5   1,182 402 38 19 26 14 12 10
6   1,193 379 38 15 24 10 14 12
female 40                
1   997 361 27 11 19 8 8 8
2   886 290 26 10 18 7 8 6
3   887 369 23 9 15 6 8 5
4   1,136 333 30 11 20 8 10 8
5   1,103 349 40 28 23 9 17 26
6   1,057 418 34 22 22 12 12 15
5-14                  
male 132                
1   1,580 511 45 20 30 11 16 19
2   1.526 451 45 16 29 10 15 11
3   1,515 514 40 14 27 10 13 8
4   1,605 508 40 17 29 14 11 9
5   1,670 491 55 22 34 14 21 17
6   1,629 500 49 17 34 12 16 13
female 125                
1   1,474 469 42 19 27 11 15 16
2   1,351 353 39 12 26 8 13 7
3   1,337 474 35 13 24 9 12 7
4   1,405 367 36 11 23 7 12 8
5   1,577 593 48 21 32 14 16 14
6   1,505 425 47 18 31 19 16 15
15-44                  
Male 181                
1   2,823 932 71 23 52 19 22 18
2   2,368 797 68 25 46 18 24 16
3   2,295 815 60 21 40 14 20 13
4   2,679 886 66 22 44 16 22 12
5   2,686 825 78 27 53 21 28 23
6   2,713 747 81 26 53 20 30 19
Female 194                
1   2,144 619 58 23 39 13 19 19
2   1,813 562 51 20 35 13 17 13
3   1,911 548 49 15 34 10 15 9
4   2,009 564 50 14 34 11 16 9
5   2,126 704 70 31 44 17 26 24
6   2,071 571 65 25 43 14 22 19
45                  
Male 60                
1   2,647 828 69 24 50 19 24 26
2   2,181 735 65 24 42 18 23 15
3   2,129 793 58 22 37 16 21 15
4   2,141 765 54 19 36 13 19 10
5   2,477 976 71 28 48 22 25 22
6   2,387 852 71 26 50 23 21 13
Female 55                
1   1,756 536 49 19 32 11 18 13
2   1,462 575 43 18 28 14 14 9
3   1,656 540 42 14 29   13 7
4   1,646 588 42 16 28 12 13 7
5   1,827 522 53 21 36 15 17 15
6   1,846 570 56 24 36 13 20 17

Coefficients of variation for within- and between-person variations of nutrient Y (CVw and CVb) were computed as

CVw = Sw/mean Y and CVb = Sb/mean Y.

Variance ratios were estimated by /

Upper and lower 95%-confidence limits on the variance ratios were approximated using where I represents the number of subjects, n the average number of repeated observations on each subject, and a is 0.05 [28; 29]. Although the confidence intervals on variance ratios will have to be interpreted with caution given the underlying non-normal data, they are used to stress that variance ratios do not overlap with 1.

The presence of seasonality was analysed by including a sinusoidal function as the fixed effect (factor Xij) in the model equation where k is an integer number that identifies the week of the year in which visit j was made. Its value ranges from 1 to 52, starting with 1 at the beginning of the study. The cyclical model was suggested by the pattern in the data in table 2 on both total energy and protein intake, and by the physical plausibility of consumption habits that parallel the cyclical behaviour of the weather and harvest.

Yij = u + subjecti + a sin(2p k/52) + b cos(2p k /52) + e ij

The ANOVA model assumes that measurements across time have a constant correlation between pairs of visits. The correlations between total energy or protein intakes on consecutive measurements or measurements taken farther apart do not differ appreciably, suggesting that our model is reasonable. The range of correlation coefficients was .52 to .61 for total energy intake, .32 to .53 for total protein, .39 to .53 for cereal protein, and .07 to .33 for protein of non-cereal origin.

Total protein and protein of cereal origin are highly correlated with total energy (calorie) intake for both sexes and all age groups on each measurement day. Correlation coefficients range between .74 and .88 for total protein and between .80 and .91 for protein of cereal origin. Proteins of non-cereal origin show a moderate but consistent correlation with total energy intake, a range of .28 to .55. Therefore, to gain information on the variability of protein intake beyond what we learn about total energy intake, protein has to be adjusted by the total energy consumed on the day of the visit. The analyses that follow were done using protein adjusted through linear regression for total energy intake on the visit day. The residuals of the linear regression were included as dependent variables in the mixed ANOVA model described above. For comparison, alternative analyses with unadjusted proteins were also conducted.

The data set was analysed using SAS statistical procedures [30]. The analysis of the within- and between-person components of variation in diet was done using the SAS VARCOMP and GLM procedures. Analyses were conducted for the entire population and also within each stratum of age and sex. Results reported are only for the age- and sex-specific analyses.

 

Results

In total, 4,431 24-hour dietary measurements were available for analysis. Table 3 shows intra-class correlation coefficients, coefficients of variation, and ratios of within- to between-person variation for total energy and protein intake, with protein intake adjusted as indicated above.

The within-person component of variation is, in all cases, greater than between-person variation. Intra-class correlation coefficients for protein intake on different days are between 0 and .14, except for men over 45 years of age.

Ratios for total energy intake range between 2.2 and 5.8 and are smaller than those shown for proteins. Intra-class correlation coefficients for total energy are also higher than for proteins.

Protein intake is relatively less consistent overall within persons than between persons compared with total energy intake. Ratios for total proteins range between 5.4 for men over 45 years of age and 215.8 for boys under 5. These ratios are due to large variability within persons and comparatively small variation between persons. Children 0-4 years old had particularly large within- and between-person coefficients of variation.

Variance ratios for cereal proteins range from 4.0 for men over 45 years old to 23.9 for boys under 5, and 93.2 for women over 45. The degree of consistency is high between individuals, with all CVb's being smaller than 10%, and 2 to 10 times smaller than CVw. Nevertheless, CVw for cereal proteins is smaller than CVw for total or non-cereal protein intakes, and smaller than that for total energy intake.

Proteins of non-cereal origin show in all cases the largest variance ratios, largest CVw, and smallest intra-class correlation coefficient. Their coefficients of variation between persons are also consistently the largest, with the exception of that for women over 45 years of age.

There are some sex differences of interest. Females have larger variance ratios in all cases than males of the same age, with the exception of children under 5 years old. This can be explained by the fact that females generally have larger CVw values than males, which more than offset generally lower values of CVb Males under 5 and females over 45 show the largest ratios, whereas males over 45 have the smallest. Age trends are not apparent.

Seasonal trends in the data were statistically significant for total energy intake (p<.0001), although they were not for total energy-adjusted protein. Nevertheless, adjustment by seasonality did not appreciably modify the estimates of within- or between-person variability. Values of ratios adjusted for seasonal changes were almost identical to those presented in table 3. Thus, adjustment for seasonal changes in diet do not alter the relative distribution of within- and between-person components of the variation.

In an alternative analysis, with results not shown, we considered variance ratios for protein intake that were not adjusted for total energy intake. These ratios for unadjusted protein intake are smaller and more similar within and across different age subgroups. This greater homogeneity can be partially understood to be a result of the large correlation be tween total energy intake and the amount of protein eaten; total amounts of food items may vary less from day to day than their relative concentration in the diet.

TABLE 3. Coefficients of variation and variance ratios for daily energy and energy adjusted protein intake by age group and sex

  N Mean Intra-class correlation coefficient Coefficients of variation (%) Variance ratio (Sw2/Sb2)
CVw CVb
0-4 years            
male 47          
energy (kcal)   1.108 .18 32 13 5.8
protein (g)            
total   32 .08 29 2 215.8
cereal   21 .05 25 5 23.9
non-cereal   11 .03 86 16 28.5
female 40          
energy (kcal)   1,005 .18 32 15 4.7
protein (g)            
total   30 .05 40 11 12.4
cereal   19 .14 24 10 5.5
non-cereal   10 .04 122 31 15.3
5-14 years            
male 132          
energy (kcal)   1,588 .31 26 17 2.2
protein (g)            
total   46 .10 25 8 9.2
cereal   31 .14 22 9 5.9
non-cereal   15 .10 78 26 9.0
female 125          
energy (kcal)   1,442 .20 29 14 4.0
protein (g)            
total   41 .05 26 6 18.7
cereal   27 .06 20 5 16.6
non-cereal   14 .07 79 21 13.6
15-44 years            
male 181          
energy (kcal)   2,594 .30 28 18 2.3
protein (g)            
total   71 .07 19 5 13.0
cereal   48 .14 19 8 6.1
non-cereal   24 .07 64 18 12.9
female 194          
energy (kcal)   2.012 .23 27 14 3.4
protein (g)            
total   57 .02 27 3 61.8
cereal   38 .14 18 7 6.2
non-cereal   19 .03 82 15 31.0
45 years            
male 60          
energy (kcal)   2,331 .31 30 20 2.2
protein (g)            
total   65 .16 18 8 5.4
cereal   44 .20 18 9 4.0
non-cereal   22 .24 69 39 3.1
female 55          
energy (kcal)   1,699 .24 29 16 3.1
protein (g)            
total   47 .00 24 0 _
cereal   32 .01 21 2 93.2
non-cereal   16 .01 70 7 109.8

Discussion

Dietary intake in developing countries is usually assumed to be homogeneous, and it has been thought that a limited number of measurements-often only one-could suffice to estimate typical individual intake over a period of time. We found that within-person variability of total energy and protein intake over one year is at least as large as that reported for industrialized countries [4;13-17; 31].

Energy-adjusted CVw and variance ratios for proteins of cereal origin are smaller than those for total protein or for proteins of non-cereal origin. Proteins of non-cereal origin have high ratios and variability in diet within individuals comparable to those reported by other authors for micronutrients and vitamins. All these findings are consistent with a diet with rice as a staple food irregularly supplemented with other food items. Low CVb also reflects a diet that does not differ much between individuals except for non-cereal proteins, which show large between-person variation.

Table 4 shows the number of measurements that would be necessary to estimate total energy-adjusted individual intake with 20% accuracy 95% of the time. Estimation of true individual intake of proteins of cereal origin may require only a few spaced measurements, about four. Estimation of total energy intake and total protein will require about seven measurements. Estimation of proteins of non-cereal origin with the same degree of accuracy will require about sixty 24-hour measurements. Thus, epidemiologic studies trying to estimate individual dietary intakes using 24-hour recall methods will have limitations. It could be expected that more measurements will be necessary to estimate true intake of micronutrients.

Although there is a seasonal trend in nutrient intake, accounting for it adds virtually nothing to the estimation of ratios of to There are two possible explanations for this. Large within-person variations may have obscured estimates of seasonal changes determined by one single measurement. Alternatively, seasonal changes may be very small in comparison to and .

The factors that determine within-person variability of diet are not well understood. Seasonality, day of the week, and random variation have been suggested. A day-of-the-week effect was not found in other parts of South Asia [32]. Although it was shown in Canadian women for some nutrients, it disappeared for most of them once nutrients were adjusted for caloric

TABLE 4. Implications for dietary measurement of within-person variability on total energy and total energy-adjusted protein intake: number of day measurements needed to estimate average dietary intake for the year within 20% of the true value 95% of the time

Sex and

age

group

(years)

Energy

Protein

Total Cereal Non -

cereal

Male
0-5 10 9 7 72
5-14 6 6 4 59
15-44 7 3 3 40
45 9 3 3 45
Female
0-4 10 16 7 143
5-14 8 7 4 60
15-44 7 7 3 64
45 8 6 4 47

Formula used: n = (Z2 * ) / (Do2 * X2),

where Z= the normal deviate for the percentage of time the measured value should be within the specified limit, = within-person variation, Xw = mean age- and sex-specific nutrient intake, Do = the percentage of long-term true intake within which it is desired that the measured value should fall. Derived from: X Z/2

Adjustment for total energy intake was made using the residuals of the regression of protein intake by total energy intake on the same visit day as the end points.

Measurement errors could be responsible for some of the large ratios found. All methods to record diet have some error [34; 35], which cannot be completely removed. A parallel estimation of diet using a different method is not available, but a pilot study conducted before the beginning of the follow-up study found a high concordance between the results using traditional weighing methods and the mixed volumetric and weighing methods used in this study. Measurement error could also have differentially affected nutrients that may be easier to record or bias the recording of food consumption outside the household. The effect of error in this study could have been to overestimate the ratio by increasing the fraction of variation attributed to within-person variability.

The present study addresses the issue of within-person variability during a one-year period for total energy and protein intakes. Nevertheless, since intra-class correlation coefficients were similar for two consecutive months and longer, within-person variability will be similar for periods shorter than a year. This conclusion is potentially important since epidemiologic studies in developing countries are often concerned with average dietary intake over shorter periods than a year.

Finally, this study found that, as in industrialized countries, a single day of dietary assessment cannot adequately characterize the diet of persons in rural Bangladesh,

 

Acknowledgements

The authors want to thank Drs. Karen Peterson, John Wyon, Nevin Scrimshaw, Doris Schopper, and an anonymous reviewer for their interest, encouragement, and constructive comments, and the ICDDR,B for its support for the original field study.

Support for this research was provided by a scholarship from the Spanish Ministry of Health and the Junta de Andalucia in Spain.

 

References

1. Rosner B. Screening for hypertension: some statistical observations. J Chron Dis 1977;30:7-18.

2. Rosner B, Polk F. The implications of blood pressure variability for clinical and screening purposes. J Chron Dis 1979;32:451-61.

3. Rosner B, Willett W. Interval estimates for correlation coefficients corrected for within-person variation: implications for study design and hypothesis testing. Am J Epid 1988;127(2):377-86.

4. Liu K, Stamler J, Dyer A, McKeever J, McKeever P. Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol. J Chron Dis 1978;31:399-418.

5. El Lozy M. Dietary variability and its impact on nutritional epidemiology. J Chron Dis 1983;36:237-49.

6. Nelson M, Black AK, Morris JA, Cole TJ. Between-and within-subject variation in nutrient intake from infancy to old age: estimating the number of days required to rank dietary intakes with desired precision. Am J Clin Nutr 1989;50:155-67.

7. Willett W. Nature of variation in diet. In: Willett W, ed. Nutritional epidemiology. New York: Oxford University Press, 1990:34-51.

8. Jacobs DR, Anderson JT, Blackburn H. Diet and serum cholesterol: do zero correlations negate the relationship? Am J Epidemiol 1979;110(1):77-87.

9. Gardner MJ, Heady JA. Some effects of within-person variability in epidemiological studies. J Chron Dis 1973;26:781 -95.

10. Hankin JH, Reynolds WE, Margen S. A short dietary method for epidemiologic studies. Am J Clin Nutr 1967;20(9):935-45.

11. Garn SM, Larkin FA, Cole PE. The real problem with l-day diet records. Am J Clin Nutr 1978;31(7): 1114-16.

12. Balogh M, Kahn HA, Medalie JH. Random repeat 24-hour dietary recalls. Am J Clin Nutr 1971 ;24:30410.

13. Sempos CT, Johnson NE, Smith EL, Gilligan G. Effects of intra-individual and inter-individual variation in repeated dietary records. Am J Epidemiol 1985;121(1): 120-30.

14. Black AK, Cole TJ, Wiles SJ, White F. Daily variation in food intake of infants from 2 to 18 months. Humn Nutr Appl Nutr 1983;37:448-58.

15. Beaton GH, Milner J, Corey P et al. Sources of variation in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr 1979;32:2546-59.

16. Beaton GH, Milner J, McGuire V, Feather TE, Little JA. Source of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation: carbohydrate sources, vitamins, and minerals. Am J Clin Nutr 1983;37:986-95.

17. Hunt WC, Leonard AG, Garry PJ, Goodwin JS. Components of variance in dietary data for an elderly population. Nutr Res 1983;3:433-44.

18. Walker AM, Blettner M. Comparing imperfect measures of exposure. Am J Epidemiol 1985;121:78390.

19. Institute of Nutrition and Food Sciences. Nutrition Survey of Rural Bangladesh, 1975-76. Dhaka. Bangladesh: University of Dhaka. 1977.

20. Chambers R, Longhurst R, Pacey A, eds. Seasonal dimensions to rural poverty. Totowa, NJ. USA: Allanheld, Osmun & Co., 1981.

21. Schaefer AK. Can nutritional status be determined from consumption or other measures? In: Assessing changing food consumption patterns. Washington, DC: National Academy Press, 1981.

22. Brown KH, Black RE, Becker S. Seasonal changes in nutritional status and the prevalence of malnutrition in a longitudinal study of young children in rural Bangladesh. Am J Clin Nutr 1982;36:303-13.

23. Chen LC, Chowdhury AKMA, Huffman SL. Seasonal dimensions of energy protein malnutrition in rural Bangladesh: the role of agriculture, dietary practices and infection. Ecol Food Nutr 1979;8: 175-87.

24. Chowdhury A, Huffman S, Chen L. Agriculture and nutrition in Matlab Thana, Bangladesh. In: Chambers R, Longhurst R, Pacey A, eds. Seasonal dimensions to rural poverty. Totowa, NJ, USA: Allanheld, Osmun & Co., 1981:52-66.

25. Van Staveren W, Deurenberg P, Burema J, Degroot LCPGM, Hautvast JGAJ. Seasonal variation in food intake, pattern of physical activity and change in body weight in a group of young adult Dutch women consuming self-selected diets. Int J Obesity 1986;111: 133-45.

26. Ziegler RG, Wilcox HB, Mason TJ, Bill JS, Virgo PW. Seasonal variation in intake of carotenoids and vegetables and fruits among white men in New Jersey. Am J Clin Nutr 1987;45:107-14.

27. Gopalan C. Nutritive value of Indian foods. New Delhi: National Institute of Nutrition, 1978.

28. Miller RG. Beyond ANOVA: basics of applied statistics. New York: John Wiley & Sons, 1986: 150-57.

29. Snedecor GW, Cochran WG. Statistical methods. 7th ed. Ames, Iowa, USA: Iowa State University Press, 1980:484-87.

30. SAS/STAT user's guide. Cary, NC, USA: SAS Institute Inc., 1988

31. Willett W, Sampson L, Stampfer M et al. Reproducibility and validity of a semi-quantitative food frequency questionaire. Am J Epidemiol 1985;122(1):51-65

32. Taskar AD, Swaminathan MC, Madhavan S. Diet surveys by weightment method. A comparison of random days, three-day and seven-day period. Ind J Med Res 1967;55(1):90-96.

33. Morgan KJ, Johnson SR, Goungetas B. Variability of food intakes: an analysis of a 12-day data series using persistence measures. Am J Epidemiol 1987;126(2):326

34. Witschi JC. Short term dietary recall and recording methods. In: Willett W, ed. Nutritional epidemiology.

New York: Oxford University Press, 1990:52-68.

35. Block G. A review of validations of dietary assessment methods. Am J Epidemiol 1982;115(4):492-505.


Contents - Previous - Next