A leading coronavirus forecasting model — used by the White House — predicted Friday that some states may be able to ease social distancing restrictions and reopen as early as May 4.

But on the same day those projections were issued by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington — giving recommended dates for reopening all 50 states — a consortium of experts in Texas released a competing model that points out what they call flaws in the IHME model and analysis showing instances in which IHME’s projections have fallen short of reality.

The clashing data and projections highlight the uncertainty U.S. leaders will face in coming months as they grapple with how to reopen the country and its economy without risking a resurgence of viral infections, overwhelmed hospitals and deaths.

The creator of IHME’s model, Christopher Murray, said his group retooled its model to be able to recommend specific dates because of intense debates in recent weeks over when and how to reopen states.

President Trump has repeatedly declared he wants states reopened as soon as possible, declaring in back-to-back tweets Friday: “LIBERATE MINNESOTA,” “LIBERATE MICHIGAN” and “LIBERATE VIRGINIA.” Also Friday, the governors of Texas and Vermont announced dates for easing certain restrictions.

The IHME model projects that at least four states could ease restrictions as early as May 4: Hawaii, Montana, Vermont and West Virginia. But other states need to wait as late as June or early July: Arkansas, Iowa, Nebraska, North Dakota, Oklahoma, South Dakota and Utah.

The new dates recommended by the IHME represent when states can shift from reliance on drastic restrictions to other strategies such as testing, contact tracing and targeted quarantines to keep the virus in check.

The key measurement IHME modelers used in determining the date is when they believe infections in a state will drop below 1 infection per 1 million people. But that measurement is imperfect because many other factors that go into deciding to lift restrictions — including whether a state has enough testing or enough public health workers to do the labor-intensive work of contact tracing. Testing capacity remains woefully lacking across the country, and local health departments lack the necessary staff, money and training to do contact tracing called for in most experts’ plans to reopen.

Murray said his IHME modelers tried to include such factors in their models but could not find data that could be used empirically. Instead, they decided to use the 1 infection per 1 million cutoff with the assumption that, at that low level, states would be more readily able to do the necessary testing and contact tracing.

Many epidemiologists and modelers have expressed growing concern over the IHME model. While it is one of the only models that offers projections state-by-state on exact dates for projected peaks and national forecasts on deaths, its projections have often clashed with other models. Its projections for equipment shortages and deaths are often below other models. The White House has at times used the model’s more-optimistic estimates to deny equipment requests.

On Friday, a few hours before the IHME announcement, a consortium at the University of Texas at Austin released a model that takes the IHME forecast as a starting point but tries to correct what the Texas researchers see as flaws.

One of the biggest flaws the Texas experts point out is that the IHME model claims more certainty as it moves further into the future, with a shrinking margin of error. That runs counter to how most models work, because the future almost always becomes increasingly uncertain in long-range forecasts. In weather forecasts, for example, predicting rain tomorrow is easier than predicting rain a month out.

“The motivation for creating our model was a concern about the certainty people may be attributing to the IHME model,” said Lauren Ancel Meyers, who led the Texas team of researchers.

Other critical differences: The IHME model predicts the United States already passed its peak of deaths this week. The Texas model takes a different approach, attaching probabilities to dates. There is only a 17 percent chance that the peak has already passed, it found, and an 80 percent chance the peak will happen by May 7.

The Texas researchers’ paper also included analysis of IHME’s past predictions compared with actual deaths per day and found, for example, that the model underestimated deaths in Italy and Spain. The number of deaths in recent days, in fact, did not fall within IHME’s projected margin of error.

Meyers said the IHME model has much value and pointed out her model would not exist if Murray did not create the IHME model.

“This is not a competition. We’re standing on their shoulders,” she said. “But like many modelers, we are continually trying to improve our methods and make our projections as reliable as possible. With cities and states making life and death decisions based on these models, the stakes are high.”

Murray similarly cautioned, “If I were a governor of a state, I would not make the decision to reopen based on just our model. We are trying to give governors and others a sense of when the risk of resurgence is going to get lower. . . . I would recommend looking at these types of models, but also a range of indicators like the capacity of your public health workers and whether your cases and deaths have fallen to a low enough level.”