for loops have their place even though they have a reputation for being slow, which can happen if interim steps aren't written to a pre-allocated object, etc. In many of the cases in which users coming from a procedural language background, the simpler functional approach does not readily come to mind.
In this case, having the indices of the elements of zz that satisfies the conditions provides a step to further processing.
I've illustrated one solution using made up data in the absence of zz. See the FAQ: How to do a minimal reproducible example reprex for beginners.
set.seed(137)
the_series <- ts(runif(120,min = -5, max = 5), frequency = 12, start = c(2010,1))
the_series[which(the_series < -2 | the_series > 2)]
#> [1] 4.141790 2.636114 4.576050 4.113563 2.720766 -2.466418 3.043643
#> [8] 3.735802 2.831292 -2.490212 3.707835 -2.280705 -2.258403 4.240154
#> [15] 4.221956 -4.472150 -2.899526 3.221924 2.189642 4.753328 -2.436642
#> [22] 3.341706 -4.841652 -3.325084 2.899013 -4.701774 -4.506472 2.873523
#> [29] -2.228279 -3.103191 3.616198 -4.472914 2.873050 4.287013 -3.440023
#> [36] -4.113404 2.982510 -3.183971 -4.010218 3.287020 -2.874006 4.324638
#> [43] 3.905566 4.635970 -2.791794 -4.778578 -4.809622 4.495908 2.608036
#> [50] -3.716669 -4.776359 2.303230 -3.892756 3.834854 -2.011916 -2.746332
#> [57] -2.250228 4.210272 3.397424 -3.849455 2.658914 -3.495754 -3.520557
#> [64] -2.944492 -4.975042 2.897202 -3.662780 4.416238 -3.892748 4.062615
#> [71] 4.954913 2.748661 3.597827 2.046638 -4.808251 -2.415291 2.955273
#> [78] 4.588577 2.076286
# the indices
which(the_series < -2 | the_series > 2)
#> [1] 3 4 6 7 8 10 12 13 14 16 17 18 19 20 23 26 28 29 31
#> [20] 32 33 34 36 38 40 41 42 44 46 47 48 49 53 54 57 58 61 62
#> [39] 64 65 66 68 69 70 72 73 74 75 78 79 80 81 85 86 87 88 89
#> [58] 90 91 92 93 94 96 97 98 99 100 103 104 105 106 107 108 112 114 115
#> [77] 118 119 120