ideal: analysis of educational testing data and roll call data with...

In pscl: Political Science Computational Laboratory

analysis of educational testing data and roll call data with IRT models, via Markov chain Monte Carlo methods

Description

Analysis of rollcall data via the spatial voting model; equivalent to a 2 parameter item-response model to educational testing data. Model fitting via Markov chain Monte Carlo (MCMC).

Usage

ideal(object, codes = object$codes,
      dropList = list(codes = "notInLegis", lop = 0),
      d = 1, maxiter = 10000, thin = 100, burnin = 5000,
      impute = FALSE,
      normalize = FALSE,
      meanzero = normalize,
      priors = NULL, startvals = "eigen",
      store.item = FALSE, file = NULL,
      verbose=FALSE, use.voter=NULL)

Arguments

object	an object of class rollcall
codes	a list describing the types of voting decisions in the roll call matrix (the votes component of the rollcall object); defaults to object$codes, the codes in the rollcall object.
dropList	a list (or alist) listing voting decisions, legislators and/or votes to be dropped from the analysis; see dropRollCall for details.
d	numeric, (small) positive integer (default = 1), dimensionality of the ability space (or "policy space" in the rollcall context).
maxiter	numeric, positive integer, multiple of thin, number of MCMC iterations
thin	numeric, positive integer, thinning interval used for recording MCMC iterations.
burnin	number of MCMC iterations to run before recording. The iteration numbered burnin will be recorded. Must be a multiple of thin.
impute	logical, whether to treat missing entries of the rollcall matrix as missing at random, sampling from the predictive density of the missing entries at each MCMC iteration.
normalize	logical, impose identification with the constraint that the ideal points have mean zero and standard deviation one, in each dimension. For one dimensional models this option is sufficient to locally identify the model parameters. See Details.
meanzero	to be deprecated/ignored; use normalize instead.
priors	a list of parameters (means and variances) specifying normal priors for the legislators' ideal points. The default is NULL, in which case the normal priors used have mean zero and precision 1 for the ideal points (ability parameters) and mean zero and precision .04 (variance 25) for the bill parameters (item discrimination and difficulty parameters). If not NULL, priors must be a list with as many as four named components xp, xpv, bp, bpv: xp a n by d matrix of prior means for the legislators' ideal points; or alternatively, a scalar, which will be replicated to fill a n by d matrix. xpv a n by d matrix of prior precisions (inverse variances); or alternatively, a scalar, which will be replicated to fill a n by d matrix. bp a m by d+1 matrix of prior means for the item parameters (with the item difficulty parameter coming last); or alternatively, a scalar, which will be replicated to fill a m by d+1 matrix. bpv a m by d+1 matrix of prior precisions for the item parameters; or alternatively, a scalar, which will be replicated to fill a m by d+1 matrix. None of the components should contain NA. If any of the four possible components are not provided, then the corresponding component of priors is assigned using the default values described above.
startvals	either a string naming a method for generating start values, valid options are "eigen" (the default), "random" or a list containing start values for legislators' ideal points and item parameters. See Details.
store.item	logical, whether item discrimination parameters should be stored. Storing item discrimination parameters can consume a large amount of memory. These need to be stored for prediction; see predict.ideal.
file	string, file to write MCMC output. Default is NULL, in which case MCMC output is stored in memory. Note that post-estimation commands like plot will not work unless MCMC output is stored in memory.
verbose	logical, default is FALSE, which generates relatively little output to the R console during execution.
use.voter	A vector of logicals of length n controlling which legislators' vote data informs item parameter estimates. Legislators corresponding to FALSE entries will not have their voting data included in updates of the item parameters. The default value of NULL will run the standard ideal-point model, which uses all legislators in updating item parameters. See Jessee (2016).

Details

The function fits a d+1 parameter item-response model to the roll call data object, so in one dimension the model reduces to the two-parameter item-response model popular in educational testing. See References.

Identification: The model parameters are not identified without the user supplying some restrictions on the model parameters; i.e., translations, rotations and re-scalings of the ideal points are observationally equivalent, via offsetting transformations of the item parameters. It is the user's responsibility to impose these identifying restrictions if desired. The following brief discussion provides some guidance.

For one-dimensional models (i.e., d=1), a simple route to identification is the normalize option, by imposing the restriction that the means of the posterior densities of the ideal points (ability parameters) have mean zero and standard deviation one, across legislators (test-takers). This normalization supplies local identification (that is, identification up to a 180 degree rotation of the recovered dimension).

Near-degenerate “spike” priors (priors with arbitrarily large precisions) or the constrain.legis option on any two legislators' ideal points ensures global identification in one dimension.

Identification in higher dimensions can be obtained by supplying fixed values for d+1 legislators' ideal points, provided the supplied fixed points span a d-dimensional space (e.g., three supplied ideal points form a triangle in d=2 dimensions), via the constrain.legis option. In this case the function defaults to vague normal priors on the unconstrained ideal points, but at each iteration the sampled ideal points are transformed back into the space of identified parameters, applying the linear transformation that maps the d+1 fixed ideal points from their sampled values to their fixed values. Alternatively, one can impose restrictions on the item parameters via constrain.items. See the examples in the documentation for the constrain.legis and constrain.items.

Another route to identification is via post-processing. That is, the user can run ideal without any identification constraints. This does not pose any formal/technical problem in a Bayesian analysis. The fact that the posterior density may have multiple modes doesn't imply that the posterior is improper or that it can't be explored via MCMC methods. – but then use the function postProcess to map the MCMC output from the space of unidentified parameters into the subspace of identified parameters. See the example in the documentation for the postProcess function.

When the normalize option is set to TRUE, an unidentified model is run, and the ideal object is post-processed with the normalize option, and then returned to the user (but again, note that the normalize option is only implemented for unidimensional models).

Start values. Start values can be supplied by the user, or generated by the function itself.

The default method, corresponding to startvals="eigen", first forms a n-by-n correlation matrix from the double-centered roll call matrix (subtracting row means, and column means, adding in the grand mean), and then extracts the first d principal components (eigenvectors), scaling the eigenvectors by the square root of their corresponding eigenvector. If the user is imposing constraints on ideal points (via constrain.legis), these constraints are applied to the corresponding elements of the start values generated from the eigen-decomposition. Then, to generate start values for the rollcall/item parameters, a series of binomial glms are estimated (with a probit link), one for each rollcall/item, j = 1, \ldots, m. The votes on the j-th rollcall/item are binary responses (presumed to be conditionally independent given each legislator's latent preference), and the (constrained or unconstrained) start values for legislators are used as predictors. The estimated coefficients from these probit models are used as start values for the item discrimination and difficulty parameters (with the intercepts from the probit GLMs multiplied by -1 so as to make those coefficients difficulty parameters).

The default eigen method generates extremely good start values for low-dimensional models fit to recent U.S. congresses, where high rates of party line voting result in excellent fits from low dimensional models. The eigen method may be computationally expensive or lead to memory errors for rollcall objects with large numbers of legislators.

The random method generates start values via iid sampling from a N(0,1) density, via rnorm, imposing any constraints that may have been supplied via constrain.legis, and then uses the probit method described above to get start values for the rollcall/item parameters.

If startvals is a list, it must contain the named components x and/or b, or named components that (uniquely) begin with the letters x and/or b. The component x must be a vector or a matrix of dimensions equal to the number of individuals (legislators) by d. If supplied, startvals$b must be a matrix with dimension number of items (votes) by d+1. The x and b components cannot contain NA. If x is not supplied when startvals is a list, then start values are generated using the default eiegn method described above, and start values for the rollcall/item parameters are regenerated using the probit method, ignoring any user-supplied values in startvals$b. That is, user-supplied values in startvals$b are only used when accompanied by a valid set of start values for the ideal points in startvals$x.

Implementation via Data Augmentation. The MCMC algorithm for this problem consists of a Gibbs sampler for the ideal points (latent traits) and item parameters, conditional on latent data y^*, generated via a data augmentation (DA) step. That is, following Albert (1992) and Albert and Chib (1993), if y_{ij} = 1 we sample from the truncated normal density

y_{ij}^* \sim N(x_i' \beta_j - \alpha_j, 1)\mathcal{I}(y_{ij}^* \geq 0)

and for y_{ij}=0 we sample

y_{ij}^* \sim N(x_i' \beta_j - \alpha_j, 1)\mathcal{I}(y_{ij}^* < 0)

where \mathcal{I} is an indicator function evaluating to one if its argument is true and zero otherwise. Given the latent y^*, the conditional distributions for x and (\beta,\alpha) are extremely simple to sample from; see the references for details.

This data-augmented Gibbs sampling strategy is easily implemented, but can sometimes require many thousands of samples in order to generate tolerable explorations of the posterior densities of the latent traits, particularly for legislators with short and/or extreme voting histories (the equivalent in the educational testing setting is a test-taker who gets almost every item right or wrong).

Value

a list of class ideal with named components

n	numeric, integer, number of legislators in the analysis, after any subsetting via processing the dropList.
m	numeric, integer, number of rollcalls in roll call matrix, after any subsetting via processing the dropList.
d	numeric, integer, number of dimensions fitted.
x	a three-dimensional array containing the MCMC output with respect to the the ideal point of each legislator in each dimension. The three-dimensional array is in iteration-legislator-dimension order. The iterations run from burnin to maxiter, at an interval of thin.
beta	a three-dimensional array containing the MCMC output for the item parameters. The three-dimensional array is in iteration-rollcall-parameter order. The iterations run from burnin to maxiter, at an interval of thin. Each rollcall has d+1 parameters, with the item-discrimination parameters stored first, in the first d components of the 3rd dimension of the beta array; the item-difficulty parameter follows in the final d+1 component of the 3rd dimension of the beta array.
xbar	a n by d matrix containing the means of the MCMC samples for the ideal point of each legislator in each dimension, using iterations burnin to maxiter, at an interval of thin.
betabar	a m by d+1 matrix containing the means of the MCMC samples for the item-specific parameters, using iterations burnin to maxiter, at an interval of thin.
args	calling arguments, evaluated in the frame calling ideal.
call	an object of class call, containing the arguments passed to ideal as unevaluated expressions or values (for functions arguments that evaluate to scalar integer or logical such as maxiter, burnin, etc).

Author(s)

Simon Jackman [e-mail], with help from Christina Maimone and Alex Tahk.

References

Albert, James. 1992. Bayesian Estimation of normal ogive item response curves using Gibbs sampling. Journal of Educational Statistics. 17:251-269.

Albert, James H. and Siddhartha Chib. 1993. Bayesian Analysis of Binary and Polychotomous Response Data. Journal of the American Statistical Association. 88:669-679.

Clinton, Joshua, Simon Jackman and Douglas Rivers. 2004. The Statistical Analysis of Roll Call Data. American Political Science Review. 98:335-370.

Jackman, Simon. 2009. Bayesian Analysis for the Social Sciences. Wiley: Hoboken, New Jersey.

Jessee, Stephen. 2016. (How) Can We Estimate the Ideology of Citizens and Political Elites on the Same Scale? American Journal of Political Science.

Patz, Richard J. and Brian W. Junker. 1999. A Straightforward Approach to Markov Chain Monte Carlo Methods for Item Response Models. Journal of Education and Behavioral Statistics. 24:146-178.

Rivers, Douglas. 2003. “Identification of Multidimensional Item-Response Models.” Typescript. Department of Political Science, Stanford University.

van Dyk, David A and Xiao-Li Meng. 2001. The art of data augmentation (with discussion). Journal of Computational and Graphical Statistics. 10(1):1-111.

See Also

rollcall, summary.ideal, plot.ideal, predict.ideal. tracex for graphical display of MCMC iterative history.

idealToMCMC converts the MCMC iterates in an ideal object to a form that can be used by the coda library.

constrain.items and constrain.legis for implementing identifying restrictions.

postProcess for imposing identifying restrictions ex post.

MCMCirt1d and MCMCirtKd in the MCMCpack package provide similar functionality to ideal.

Examples

## Not run: 
## long run, many iterations
data(s109)
n <- dim(s109$legis.data)[1]
x0 <- rep(0,n)
x0[s109$legis.data$party=="D"] <- -1
x0[s109$legis.data$party=="R"] <- 1

id1 <- ideal(s109,
             d=1,
             startvals=list(x=x0),
             normalize=TRUE,
             store.item=TRUE,
             maxiter=260E3,
             burnin=10E3,
             thin=100)  

## End(Not run)

pscl documentation built on May 29, 2024, 9:09 a.m.