I want to estimate, graph, and interpret the effects of nonlinear models with interactions of continuous and discrete variables. The results I am after are not trivial, but obtaining what I want using margins, marginsplot, and factor-variable notation is straightforward. Read more…

Overview

In the first part of this post, I discussed the multinomial probit model from a random utility model perspective. In this part, we will have a closer look at how to interpret our estimation results.

How do we interpret our estimation results?

We created a fictitious dataset of individuals who were presented a set of three health insurance plans (Sickmaster, Allgood, and Cowboy Health). We pretended to have a random sample of 20- to 60-year-old persons who were asked Read more…

\(\newcommand{\xb}{{\bf x}}
\newcommand{\betab}{\boldsymbol{\beta}}
\newcommand{\zb}{{\bf z}}
\newcommand{\gammab}{\boldsymbol{\gamma}}\)We have no choice but to choose

We make choices every day, and often these choices are made among a finite number of potential alternatives. For example, do we take the car or ride a bike to get to work? Will we have dinner at home or eat out, and if we eat out, where do we go? Scientists, marketing analysts, or political consultants, to name a few, wish to find out why people choose what they choose.

In this post, Read more…

\(\newcommand{\mub}{{\boldsymbol{\mu}}}
\newcommand{\eb}{{\boldsymbol{e}}}
\newcommand{\betab}{\boldsymbol{\beta}}\)Determining the stationarity of a time series is a key step before embarking on any analysis. The statistical properties of most estimators in time series rely on the data being (weakly) stationary. Loosely speaking, a weakly stationary process is characterized by a time-invariant mean, variance, and autocovariance.

In most observed series, however, the presence of a trend component results in the series being nonstationary. Furthermore, the trend can be either deterministic or stochastic, depending on which appropriate transformations must be applied to obtain a stationary series. For example, a stochastic trend, or commonly known as a unit root, is eliminated by differencing the series. However, differencing a series that in fact contains a deterministic trend results in a unit root in the moving-average process. Similarly, subtracting a deterministic trend from a series that in fact contains a stochastic trend does not render a stationary series. Hence, it is important to identify whether nonstationarity is due to a deterministic or a stochastic trend before applying the proper transformations.

In this post, Read more…

We continue with the series of posts where we illustrate how to obtain correct standard errors and marginal effects for models with multiple steps. In this post, we estimate the marginal effects and standard errors for a hurdle model with two hurdles and a lognormal outcome using mlexp. mlexp allows us to estimate parameters for multiequation models using maximum likelihood. In the last post (Multiple equation models: Estimation and marginal effects using gsem), we used gsem to estimate marginal effects and standard errors for a hurdle model with two hurdles and an exponential mean outcome.

We exploit the fact that the hurdle-model likelihood is separable and the joint log likelihood is the sum of the individual hurdle and outcome log likelihoods. We estimate the parameters of each hurdle and the outcome separately to get initial values. Then, we use mlexp to estimate the parameters of the model and margins to obtain marginal effects. Read more…

Starting point: A hurdle model with multiple hurdles

In a sequence of posts, we are going to illustrate how to obtain correct standard errors and marginal effects for models with multiple steps.

Our inspiration for this post is an old Statalist inquiry about how to obtain marginal effects for a hurdle model with more than one hurdle (http://www.statalist.org/forums/forum/general-stata-discussion/general/1337504-estimating-marginal-effect-for-triple-hurdle-model). Hurdle models have the appealing property that their likelihood is separable. Each hurdle has its own likelihood and regressors. You can estimate each one of these hurdles separately to obtain point estimates. However, you cannot get standard errors or marginal effects this way.

In this post, Read more…

\(\newcommand{\mub}{{\boldsymbol{\mu}}}
\newcommand{\eb}{{\boldsymbol{e}}}
\newcommand{\betab}{\boldsymbol{\beta}}\)Applied time-series researchers often want to compare the accuracy of a pair of competing forecasts. A popular statistic for forecast comparison is the mean squared forecast error (MSFE), a smaller value of which implies a better forecast. However, a formal test, such as Diebold and Mariano (1995), distinguishes whether the superiority of one forecast is statistically significant or is simply due to sampling variability.

A related test is the forecast encompassing test. This test is used to determine whether one of the forecasts encompasses all the relevant information from the other. The resulting test statistic may lead a researcher to either combine the two forecasts or drop the forecast that contains no additional information.

In this post, Read more…

As of Stata 16, see [BAYES] bayesstats grubin and Bayesian analysis: Gelman-Rubin convergence diagnostic.

The original blog posted May 26, 2016, omitted option initrandom from the bayesmh command. The code and the text of the blog entry were updated on August 9, 2018, to reflect this.

Overview

MCMC algorithms used for simulating posterior distributions are indispensable tools in Bayesian analysis. A major consideration in MCMC simulations is that of convergence. Has the simulated Markov chain fully explored the target posterior distribution so far, or do we need longer simulations? A common approach in assessing MCMC convergence is based on running and analyzing the difference between multiple chains.

For a given Bayesian model, bayesmh is capable of producing multiple Markov chains with randomly dispersed initial values by using the initrandom option, available as of the update on 19 May 2016. In this post, I demonstrate the Gelman–Rubin diagnostic as a more formal test for convergence using multiple chains. For graphical diagnostics, see Graphical diagnostics using multiple chains in [BAYES] bayesmh for more details. To compute the Gelman–Rubin diagnostic, I use an unofficial command, grubin, which can be installed by typing the following in Stata: Read more…

Initial thoughts

Estimating causal relationships from data is one of the fundamental endeavors of researchers. Ideally, we could conduct a controlled experiment to estimate causal relations. However, conducting a controlled experiment may be infeasible. For example, education researchers cannot randomize education attainment and they must learn from observational data.

In the absence of experimental data, we construct models to capture the relevant features of the causal relationship we have an interest in, using observational data. Models are successful if the features we did not include can be ignored without affecting our ability to ascertain the causal relationship we are interested in. Sometimes, however, ignoring some features of reality results in models that yield relationships that cannot be interpreted causally. In a regression framework, depending on our discipline or our research question, we give a different name to this phenomenon: endogeneity, omitted confounders, omitted variable bias, simultaneity bias, selection bias, etc.

Below I show how we can understand many of these problems in a unified regression framework and use simulated data to illustrate how they affect estimation and inference. Read more…

Autoregressive (AR) and moving-average (MA) models are combined to obtain ARMA models. The parameters of an ARMA model are typically estimated by maximizing a likelihood function assuming independently and identically distributed Gaussian errors. This is a rather strict assumption. If the underlying distribution of the error is nonnormal, does maximum likelihood estimation still work? The short answer is yes under certain regularity conditions and the estimator is known as the quasi-maximum likelihood estimator (QMLE) (White 1982).

In this post, I use Monte Carlo Simulations (MCS) to verify that the QMLE of a stationary and invertible ARMA model is consistent and asymptotically normal. See Yao and Brockwell (2006) for a formal proof. For an overview of performing MCS in Stata, refer to Monte Carlo simulations using Stata. Also see A simulation-based explanation of consistency and asymptotic normality for a discussion of performing such an exercise in Stata.

Simulation

Let’s begin by Read more…