Poststratification
Stratification is a technique developed for survey sampling in which a population is partitioned into subgroups (i.e., stratified) and each group (i.e., stratum) is sampled independently. If the subgroups are more homogeneous (i.e., lower variance) than the population as a whole, this can reduce variance in the estimate of a quantity of interest at the population level.
Poststratification is a technique for adjusting a non-representative sample (i.e., a convenience sample or other observational data) for which there are demographic predictors characterizing the strata. It is carried out after a model is fit to the observed data, hence the name poststratification (Little 1993). Poststratification can be fruitfully combined with regression modeling (or more general parametric modeling), which provides estimates based on combinations of predictors (or general parameters) rather than raw counts in each stratum. Multilevel modeling is useful in determining how much partial pooling to apply in the regressions, leading to the popularity of the combination of multilevel regression and poststratification (MRP) (Park, Gelman, and Bafumi 2004).
Some examples
Earth science
Stratification and poststratification can be applied to many applications beyond survey sampling (Kennedy and Gelman 2019). For example, large-scale whole-earth soil-carbon models are fit with parametric models of how soil-carbon depends on features of an area such as soil composition, flora, fauna, temperature, humidity, etc. Given a model that predicts soil-carbon concentration given these features, a whole-earth model can be created by stratifying the earth into a grid of say 10km by 10km “squares” (they can’t literally be square because the earth’s surface is topologically a sphere). Each grid area has an estimated makeup of soil type, forestation, climate, etc. The global level of soil carbon is then estimated using poststratification by simply summing the expected soil carbon estimated for each square in the grid (Paustian et al. 1997). Dynamic models can then be constructed by layering a time-series component, varying the poststratification predictors over time, or both (Field et al. 1998).
Polling
Suppose a university’s administration would like to estimate the support for a given proposal among its students. A poll is carried out in which 490 respondents are undergraduates, 112 are graduate students, and 47 are continuing education students. Now suppose that support for the issue among the poll respondents is is 25% among undergraduate students (subgroup 1), 40% among graduate students (subgroup 2), and 80% among continuing education students (subgroup 3). Now suppose that the student body is made up of 20,000 undergraduates, 5,000 graduate students, and 2,000 continuing education students. It is important that our subgroups are exclusive and exhaustive, i.e., they form a partition of the population.
The proportion of support in the poll among students in each group provides a simple maximum likelihood estimate \(\theta^* = (0.25, 0.5, 0.8)\) of support in each group for a simple Bernoulli model where student \(n\)’s vote is modeled as \[ y_n \sim \textrm{bernoulli}(\theta_{jj[n]}), \] where \(jj[n] \in 1:3\) is the subgroup to which the \(n\)-th student belongs.
An estimate of the population prevalence of support for the issue among students can be constructed by simply multiplying estimated support in each group by the size of each group. Letting \(N = (20\,000,\, 5\,000,\, 2\,000)\) be the subgroup sizes, the poststratified estimate of support in the population \(\phi^*\) is estimated by \[ \phi^* = \frac{\displaystyle \sum_{j = 1}^3 \theta_j^* \cdot N_j} {\displaystyle \sum_{j = 1}^3 N_j}. \] Plugging in our estimates and population counts yields \[\begin{eqnarray*} \phi* & = & \frac{0.25 \cdot 20\,000 + 0.4 \cdot 5\,000 + 0.8 \cdot 2\,000} {20\,000 + 5\,000 + 2\,000} \\[4pt] & = & \frac{8\,600}{27\,000} \\[4pt] & \approx & 0.32. \end{eqnarray*}\]
Bayesian poststratification
Considering the same polling data from the previous section in a Bayesian setting, the uncertainty in the estimation of subgroup support is pushed through predictive inference in order to get some idea of the uncertainty of estimated support. Continuing the example of the previous section, the likelihood remains the same, \[ y_n \sim \textrm{bernoulli}(\theta_{jj[n]}), \] where \(jj[n] \in 1:J\) is the group to which item \(n\) belongs and \(\theta_j\) is the proportion of support in group \(j\).
This can be reformulated from a Bernoulli model to a binomial model in the usual way. Letting \(A_j\) be the number of respondents in group \(j\) and \(a_j\) be the number of positive responses in group \(j\), the likelihood may be reduced to the form \[ a_j \sim \textrm{binomial}(A_j, \theta_j). \] A simple uniform prior on the proportion of support in each group completes the model, \[ \theta_j \sim \textrm{beta}(1, 1). \] A more informative prior could be used if there is prior information available about support among the student body.
Using sampling, draws \(\theta^{(m)} \sim p(\theta \mid y)\) from the posterior may be combined with the population sizes \(N\) to estimate \(\phi\), the proportion of support in the population, \[ \phi^{(m)} = \frac{\displaystyle \sum_{j = 1}^J \theta_j^{(m)} \cdot N_j} {\displaystyle \sum_{j = 1}^J N_j}. \] The posterior draws for \(\phi^{(m)}\) characterize expected support for the issue in the entire population. These draws may be used to estimate expected support (the average of the \(\phi^{(m)}\)), posterior intervals (quantiles of the \(\phi^{(m)}\)), or to plot a histogram.
Poststratification in Stan
The maximum likelihood and Bayesian estimates can be handled with the same Stan program. The model of individual votes is collapsed to a binomial, where \(A_j\) is the number of voters from group \(j\), \(a_j\) is the number of positive responses from group \(j\), and \(N_j\) is the size of group \(j\) in the population.
data {
int<lower=1> J;
array[J] int<lower=0> A;
array[J] int<lower=0> a;
vector<lower=0>[J] N;
}
parameters {
vector<lower=0, upper=1>[J] theta;
}
model {
a ~ binomial(A, theta);
}
generated quantities {t
real<lower=0, upper=1> phi = dot(N, theta) / sum(N);
}The likelihood is vectorized, and implicitly sums over the \(j\). The prior is implicitly uniform on \((0, 1),\) the support of \(\theta.\) The summation is computed using a dot product and the sum function, which is why N was declared as a vector rather than as an array of integers.
Regression and poststratification
In applications to polling, there are often numerous demographic features like age, gender, income, education, state of residence, etc. If each of these demographic features induces a partition on the population, then their product also induces a partition on the population. Often sources such as the census have matching (or at least matchable) demographic data; otherwise it must be estimated.
The problem facing poststratification by demographic feature is that the number of strata increases exponentially as a function of the number of features. For instance, 4 age brackets, 2 sexes, 5 income brackets, and 50 states of residence leads to \(5 \cdot 2 \cdot 5 \cdot 50 = 2000\) strata. Adding another 5-way distinction, say for education level, leads to 10,000 strata. A simple model like the one in the previous section that takes an independent parameter \(\theta_j\) for support in each stratum is unworkable in that many groups will have zero respondents and almost all groups will have very few respondents.
A practical approach to overcoming the problem of low data size per stratum is to use a regression model. Each demographic feature will require a regression coefficient for each of its subgroups, but now the parameters add to rather than multiply the total number of parameters. For example, with 4 age brackets, 2 sexes, 5 income brackets, and 50 states of residence, there are only \(4 + 2 + 5 + 50 = 61\) regression coefficients to estimate. Now suppose that item \(n\) has demographic features \(\textrm{age}_n \in 1:5\), \(\textrm{sex}_n \in 1:2\), \(\textrm{income}_n \in 1:5,\) and \(\textrm{state}_n \in 1:50\). A logistic regression may be formulated as \[ y_n \sim \textrm{bernoulli}(\textrm{logit}^{-1}( \alpha + \beta_{\textrm{age}[n]} + \gamma_{\textrm{sex}[n]} + \delta_{\textrm{income}[n]} + \epsilon_{\textrm{state}[n]} )), \] where \(\textrm{age}[n]\) is the age of the \(n\)-th respondent, \(\textrm{sex}[n]\) is their sex, \(\textrm{income}[n]\) their income and \(\textrm{state}[n]\) their state of residence. These coefficients can be assigned priors, resulting in a Bayesian regression model.
To poststratify the results, the population size for each combination of predictors must still be known. Then the population estimate is constructed as \[ \sum_{i = 1}^5 \sum_{j = 1}^2 \sum_{k = 1}^5 \sum_{m = 1}^{50} \textrm{logit}^{-1}(\alpha + \beta_i + \gamma_j + \delta_k + \eta_m) \cdot \textrm{pop}_{i, j, k, m}, \] where \(\textrm{pop}_{i, j, k, m}\) is the size of the subpopulation with age \(i\), sex \(j\), income level \(k\), and state of residence \(m\).
As formulated, it should be clear that any kind of prediction could be used as a basis for poststratification. For example, a Gaussian process or neural network could be used to produce a non-parametric model of outcomes \(y\) given predictors \(x\).
Multilevel regression and poststratification
With large numbers of demographic features, each cell may have very few items in it with which to estimate regression coefficients. For example, even in a national-level poll of 10,000 respondents, if they are divided by the 50 states, that’s only 200 respondents per state on average. When data sizes are small, parameter estimation can be stabilized and sharpened by providing hierarchical priors. With hierarchical priors, the data determines the amount of partial pooling among the groups. The only drawback is that if the number of groups is small, it can be hard to fit these models without strong hyperpriors.
The model introduced in the previous section had likelihood \[ y_n \sim \textrm{bernoulli}(\textrm{logit}^{-1}( \alpha + \beta_{\textrm{age}[n]} + \gamma_{\textrm{sex}[n]} + \delta_{\textrm{income}[n]} + \epsilon_{\textrm{state}[n]} )). \] The overall intercept can be given a broad fixed prior, \[ \alpha \sim \textrm{normal}(0, 5). \] The other regression parameters can be given hierarchical priors, \[\begin{eqnarray*} \beta_{1:4} & \sim & \textrm{normal}(0, \sigma^{\beta}) \\[2pt] \gamma_{1:2} & \sim & \textrm{normal}(0, \sigma^{\gamma}) \\[2pt] \delta_{1:5} & \sim & \textrm{normal}(0, \sigma^{\delta}) \\[2pt] \epsilon_{1:50} & \sim & \textrm{normal}(0, \sigma^{\epsilon}) \end{eqnarray*}\]
The hyperparameters for scale of variation within a group can be given simple standard hyperpriors, \[ \sigma^{\beta}, \sigma^{\gamma}, \sigma^{\delta}, \sigma^{\epsilon} \sim \textrm{normal}(0, 1). \] The scales of these fixed hyperpriors need to be determined on a problem-by-problem basis, though ideally they will be close to standard (mean zero, unit variance).
Dealing with small partitions and non-identifiability
The multilevel structure of the models used for multilevel regression and poststratification consist of a sum of intercepts that vary by demographic feature. This immediately introduces non-identifiability. A constant added to each state coefficient and subtracted from each age coefficient leads to exactly the same likelihood.
This is non-identifiability that is only mitigated by the (hierarchical) priors. When demographic partitions are small, as they are with several categories in the example, it can be more computationally tractable to enforce a sum-to-zero constraint on the coefficients. Other values than zero will by necessity be absorbed into the intercept, which is why it typically gets a broader prior even with standardized data. With a sum to zero constraint, coefficients for binary features will be negations of each other. For example, because there are only two sex categories, \(\gamma_2 = -\gamma_1.\)
To implement sum-to-zero constraints,
parameters {
vector[K - 1] alpha_raw;
// ...
}
transformed parameters {
vector<multiplier=sigma_alpha>[K] alpha
= append_row(alpha_raw, -sum(alpha_raw));
// ...
}
model {
alpha ~ normal(0, sigma_alpha);
}This prior is hard to interpret in that there are K normal distributions, but only K - 1 free parameters. An alternative is to put the prior only on alpha_raw, but that is also difficult to interpret.
Soft constraints can be more computationally tractable. They are also simpler to implement.
parameters {
vector<multiplier=alpha>[K] alpha;
// ...
}
model {
alpha ~ normal(0, sigma_alpha);
sum(alpha) ~ normal(0, 0.001);
}This leaves the regular prior, but adds a second prior that concentrates the sum near zero. The scale of the second prior will need to be established on a problem and data-set specific basis so that it doesn’t shrink the estimates beyond the shrinkage of the hierarchical scale parameters.
Note that in the hierarchical model, the values of the coefficients when there are only two coefficients should be the same absolute value but opposite signs. Any other difference could be combined into the overall intercept \(\alpha.\) Even with a wide prior on the intercept, the hyperprior on \(\sigma^{\gamma}\) may not be strong enough to enforce that, leading to a weak form non-identifiability in the posterior. Enforcing a (hard or soft) sum-to-zero constraint can help mitigate non-identifiability. Whatever prior is chosen, prior predictive checks can help diagnose problems with it.
None of this work to manage identifiability in multilevel regressions has anything to do with the poststratification; it’s just required to fit a large multilevel regression with multiple discrete categories. Having multiple intercepts always leads to weak non-identifiability, even with the priors on the intercepts all centered at zero.
Coding MRP in Stan
Multilevel regression and poststratification can be coded directly in Stan. To code the non-centered parameterization for each coefficient, which will be required for sampling efficiency, the multiplier transform is used on each of the parameters. The combination of
vector<multiplier=s>[K] a;
// ...
a ~ normal(0, s);implements a non-centered parameterization for a; a centered parameterization would drop the multiplier specification. The prior scale s is being centered here. The prior location is fixed to zero in multilevel regressions because there is an overall intercept; introducing a location parameters in the prior would introduce non-identifiability with the overall intercept. The centered parameterization drops the multiplier.
Here is the full Stan model, which performs poststratification in the generated quantities using population sizes made available through data variable P.
data {
int<lower=0> N;
array[N] int<lower=1, upper=4> age;
array[N] int<lower=1, upper=5> income;
array[N] int<lower=1, upper=50> state;
array[N] int<lower=0> y;
array[4, 5, 50] int<lower=0> P;
}
parameters {
real alpha;
real<lower=0> sigma_beta;
vector<multiplier=sigma_beta>[4] beta;
real<lower=0> sigma_gamma;
vector<multiplier=sigma_gamma>[5] gamma;
real<lower=0> sigma_delta;
vector<multiplier=sigma_delta>[50] delta;
}
model {
y ~ bernoulli_logit(alpha + beta[age] + gamma[income] + delta[state]);
alpha ~ normal(0, 2);
beta ~ normal(0, sigma_beta);
gamma ~ normal(0, sigma_gamma);
delta ~ normal(0, sigma_delta);
{ sigma_beta, sigma_gamma, sigma_delta } ~ normal(0, 1);
}
generated quantities {
real expect_pos = 0;
int total = 0;
for (b in 1:4) {
for (c in 1:5) {
for (d in 1:50) {
total += P[b, c, d];
expect_pos
+= P[b, c, d]
* inv_logit(alpha + beta[b] + gamma[c] + delta[d]);
}
}
}
real<lower=0, upper=1> phi = expect_pos / total;
}Unlike in posterior predictive inference aimed at uncertainty, there is no need to introduce binomial sampling uncertainty into the estimate of expected positive votes. Instead, generated quantities are computed as expectations. In general, it is more efficient to work in expectation if possible (the Rao-Blackwell theorem says it’s at least as efficient to work in expectation, but in practice, it can be much much more efficient, especially for discrete quantities).
Binomial coding
In some cases, it can be more efficient to break the data down by group. Suppose there are \(4 \times 5 \times 2 \times 50 = 2000\) groups. The data can be broken down into a size-2000 array, with entries corresponding to total vote counts in that group
int<lower=0> G;
array[G] int<lower=1, upper=4> age;
array[G] int<lower=1, upper=5> income;
array[G] int<lower=1, upper=50> state;Then the number of positive votes and the number of total votes are collected into two parallel arrays indexed by group.
array[G] int<lower=0> pos_votes;
array[G] int<lower=0> total_votes;Finally, the likelihood is converted to binomial.
pos_votes ~ binomial_logit(total_votes,
alpha + beta[age] + ...);The predictors look the same because of the way the age and other data items are coded.
Coding binary groups
In this first model, sex is not included as a predictor. With only two categories, it needs to be modeled separately, because it is not feasible to build a hierarchical model with only two cases. A sex predictor is straightforward to add to the data block; it takes on values 1 or 2 for each of the N data points.
array[N] int<lower=1, upper=2> sex;Then add a single regression coefficient as a parameter,
real epsilon;In the log odds calculation, introduce a new term
[epsilon, -epsilon][sex]';That is, the likelihood will now look like
y ~ bernoulli_logit(alpha + beta[age] + gamma[income] + delta[state]
+ [epsilon, -epsilon][sex]');For data point n, the expression [epsilon, -epsilon][sex] takes on value [epsilon, -epsilon][sex][n], which with Stan’s multi-indexing reduces to [epsilon, -epsilon][sex[n]]. This term evaluates to epsilon if sex[n] is 1 and to -epsilon if sex[n] is 2. The result is effectively a sum-to-zero constraint on two sex coefficients. The ' at the end transposes [epsilon, -epsilon][sex] which is a row_vector into a vector that can be added to the other variables.
Finally, a prior is needed for the coefficient in the model block,
epsilon ~ normal(0, 2);As with other priors in multilevel models, the posterior for epsilon should be investigated to make sure it is not unrealistically wide.
Adding group-level predictors
If there are group-level predictors, such as average income in a state, or vote share in a previous election, these may be used as predictors in the regression. They will not pose an obstacle to poststratification because they are at the group level. For example, suppose the average income level in the state is available as the data variable
array[50] real<lower=0> income;then a regression coefficient psi can be added for the effect of average state income,
real psi;with a fixed prior,
psi ~ normal(0, 2);This prior assumes the income predictor has been standardized. Finally, a term is added to the regression for the fixed predictor,
y ~ bernoulli_logit(alpha + beta[age] + ... + delta[state]
+ income[state] * psi);And finally, the formula in the generated quantities block is also updated,
expect_pos
+= P[b, c, d]
* inv_logit(alpha + beta[b] + gamma[c] + delta[d]
+ income[d] * psi);Here d is the loop variable looping over states. This ensures that the poststratification formula matches the likelihood formula.