Estimating the size of the Universe has never been easy; the Universe being approximately 13.7 billion years old, many objects are too far away to be visible (their light did not have enough time to reach us). A team of researchers has recently used the Bayesian model averaging, a form of statistical modeling, to constrain the curvature and size of the Universe.
At first glance, as the Universe is 13.7 billion years old, it seems natural to think that any object at a greater distance than 13.7 billion light-years will not be visible. However, as the Universe is expanding, photons in the cosmic microwave background (the first light emitted) have traveled approximately 45 billion light-years to get there: this makes the observable Universe about 90 billion light-years across.
But the Universe is actually even bigger. So, how do we estimate its size having only some partial information? Fortunately, knowing the structure, i.e. the shape, of the observable Universe, allows to estimate its size. Theoretically, there are three possible structures, depending on the curvature of the Universe: closed, open and flat (see The Shape of the Universe for further information). In the two latter cases, the Universe must be infinite, but a closed shape implies the Universe has a definite volume.
There are various ways to measure the curvature of the Universe; unfortunately, the results are different, depending on the method chosen, so which to choose?
Mihran Vardanyan at the University of Oxford and his colleagues have recently presented a way to average the results given by all the data. They used a procedure called Bayesian model averaging, which estimates how likely given models are to be correct considering the available data. This has the advantage to constrain the curvature and size of the Universe much better than with other methods.
According to their results, in all likelihood, the Universe is flat, which makes its size infinite. Anyway, their calculations give a lower limit for its size: the Universe is at least 250 times bigger than what we can see.
Bayesian model averaging gives the most general parameter constraints, and Vardanyan and his colleagues believe that the formalism they presented can be employed successfully in a large variety of cosmological problems.
