Now that I reconsider, using only 4 drives makes my statement incorrect. They are both equally likely to fail. The advantage of RAID1+0 comes when you have more than 4 disks. I'll still explain why, though.

Let me just state for the record that a mirror (RAID1) can survive with only one of its submirrors, but a stripe (RAID0) must contain all of its parts.

Let's first imagine a RAID1+0 setup. Let's say that there are six drives total. There are three mirrors, each with two submirrors. Let's call the mirrors A, B, and C and their respective submirrors 1 and 2. So the drives are labelled A1, A2, B1, B2, C1, and C2. If A1 fails, the A mirror is still active. Let's call that the initial failure state. If we then lose A2, then the RAID fails because mirror A fails and the stripe needs both parts of the stripe. But if we lose B1, B2, C1, or C2 then the RAID set remains active, because the mirrors, while most are in a degraded state, are all still active. So once in its initial degraded state, the chances of one other failure taking down the drive are 1 in 5. After that, 1 in 2. After that, 1 in 1.

Let's now imagine a RAID0+1 setup. There are still six drives, but now we have two striped sets, each with three drives, that are mirrored together. Let's call the striped sets A and B and their respective internal drives 1, 2, and 3. So we have drives A1, A2, A3, B1, B2, and B3. If A1 fails, the A stripe is disabled. It no longer has the data necessary to function. If we now lose any drive in the B stripe, the B stripe goes down and the RAID ceases to function. So after its initial degraded state, the chances of another single drive failure causing the RAID to go offline are 3 in 5. After that, 3 in 4. After that, 1 in 1.

So it's a statstics game, but it's not even close. And the odds become even further separated as you add more drives.