An interactive dashboard measuring dietary diversity using the Shannon Index — in both gram-based and ratio-based formulations — tracking how evenly diet is distributed across food sources, by state, demographic group, and income level.
The Indian Council of Medical Research — National Institute of Nutrition (ICMR-NIN) publishes Dietary Guidelines for Indians, specifying the food groups and quantities needed for a nutritionally adequate diet. The dashboard uses the ICMR-NIN recommended diet for a reference adult woman (55 kg, moderate activity, non-lactating) as the benchmark.
Whole grains like rice, wheat, ragi, bajra and jowar — the energy base of the Indian diet.
Dals, legumes, eggs, fish or meat — essential sources of protein, iron and B-vitamins.
Spinach, amaranth, fenugreek and other greens rich in folate, iron and calcium.
A variety of seasonal vegetables for vitamins, minerals, fibre and antioxidants.
Carrots, radish and other root vegetables providing energy and micronutrients.
Fresh, whole fruits daily for vitamins, minerals, dietary fibre and natural sugars.
Milk and milk products for calcium, high-quality protein and essential fatty acids.
Cooking oils and visible fats used in moderation to meet essential fatty acid needs.
Groundnuts, almonds, sesame and other oilseeds for healthy fats, protein and minerals.
Borrowed from ecology and information theory, the Shannon Index captures both the richness and evenness of a distribution — here applied to how micronutrient intake is spread across food groups.
The SDI goes beyond simple calorie counts to measure how balanced a household's diet is. Two complementary formulations are used, each constructing the index at the household level.
Gram-Based Construction
Shannon Entropy (H): Household food intake across 9 food groups is first converted to per-adult-female-equivalent (AFE) quantities, then capped at ICMR-NIN requirements — so eating more than required doesn't inflate "diversity." The capped quantities are converted to shares, and Shannon entropy is computed: H = −Σ(pᵢ × log pᵢ). Higher H means intake is more evenly spread.
Adequacy Adjustment (A): For each food group, the ratio of actual intake to the ICMR-NIN requirement is computed (capped at 1). The average across all 9 groups gives an adequacy score A ∈ [0, 1] that penalises diets which may be diverse but are insufficient in quantity.
Cereal Penalty (C): If cereals' share of capped intake exceeds their share in the recommended diet, an exponential penalty C = exp(−3 × excess) is applied. This penalises cereal over-reliance.
The final gram-based Requirement-Adjusted SDI is Hadj = H × A × C, ranging from 0 to approximately 1.97. A version without the cereal adjustment (H × A) is also reported.
Ratio-Based Construction
The gram-based formulation can be dominated by food groups with large gram requirements (e.g., cereals at ~280 g/day). The ratio-based alternative corrects this by normalising each food group's intake to its requirement before computing shares, placing all groups on a common 0–1 scale:
Each food group's adequacy ratio aᵢ = min(qᵢ / rᵢ, 1) is computed, then shares are derived from these ratios: pᵢ = aᵢ / Σaᵢ. Shannon entropy is calculated on these ratio-based shares, and the adequacy score A is simply the mean of the capped ratios. The final index is Hadjratio = H × A, with a maximum of ln(9) ≈ 2.20 when all food groups are fully met.
Because the ratio-based approach puts all food groups on the same scale, the cereal penalty is unnecessary — cereal share only becomes large when other groups are genuinely under-consumed.
The dashboard provides four complementary ways to quantify dietary diversity — two formulations of the Shannon Index plus supporting metrics.
The original formulation: intake is capped at ICMR-NIN requirements in grams, shares are computed from capped gram quantities, and Shannon entropy measures evenness. Includes a cereal penalty (H × A × C) to correct for cereal dominance. Maximum ≈ 1.97.
The alternative formulation: each food group's intake is normalised to its requirement (0–1 scale) before computing shares. All food groups contribute equally when fully met, so no cereal penalty is needed (H × A). Maximum = ln(9) ≈ 2.20.
The combined SDI — Shannon entropy multiplied by the adequacy score — available for both gram-based and ratio-based formulations. Penalises diets that are diverse but insufficient in quantity.
The percentage by which cereals' share of micronutrient intake exceeds what a balanced diet would imply — a direct measure of cereal dependence in the gram-based formulation.
A fully interactive tool for researchers, policymakers, and journalists to investigate dietary diversity patterns across India.
Toggle between HCES 2011–12 and 2023–24, or use Compare mode to overlay both rounds and track how dietary diversity has changed over a decade.
Slice data by sector, state, religion, social group, household head, children, and month of survey to reveal diversity gaps across populations.
Switch between expenditure deciles and continuous MPCE (log scale) on the x-axis to see how diversity changes across the income distribution.
All estimates come from a Bayesian model. Toggle 95% credible intervals on or off for both line charts and bar comparisons.
Export charts as PNG or SVG, download the current data view as CSV, or grab the full dataset for your own analysis.
Select multiple groups simultaneously to compare diversity patterns side by side on the same chart — states, religions, castes, or sectors.
Select one or more groups within each dimension to compare diversity patterns side by side.
Rural vs. Urban — how urbanisation shapes dietary diversity.
All states and union territories, revealing vast geographic variation in diet composition.
Hindu, Muslim, Christian, Sikh, and other religious groups with distinct food cultures.
Scheduled Tribes, Scheduled Castes, OBCs, and other groups.
Male- vs. female-headed households and their differing diversity profiles.
Households with and without children — does family composition affect diversity?
Seasonal variation in dietary diversity across the survey year.
Rigorous statistical modelling underpins every number in the dashboard.
Food consumption quantities from HCES are converted into per-adult-female-equivalent (AFE) intake across 9 food groups, creating a distribution of nutrient sources for each household.
Intake is capped at ICMR-NIN requirements in grams, shares are computed from capped quantities, and Shannon entropy is calculated. A cereal penalty addresses the mechanical dominance of cereals in gram shares.
Each food group's intake is normalised to its requirement (aᵢ = min(qᵢ/rᵢ, 1)), placing all groups on a common 0–1 scale before computing shares and entropy. The cereal penalty is unnecessary in this formulation.
Both formulations are normalised, logit-transformed, and modelled with Bayesian GAMs across expenditure deciles and demographic groups, providing posterior means and 95% credible intervals.
Dive into the interactive dashboard to explore both gram-based and ratio-based diversity measures, compare groups, and download data for your own research.
Launch Dashboard