(a) Do an additive time series model on the tornado data using years 1998 through Dec. 2018. The months of 2019 will be used for a holdout sample Make sure that the assumptions are valid. Interpret your coefficients and findings.
(b) Find a multiplicative model on the same data. Use the same 12 months as a holdout sample. Make sure that assumptions are valid. Interpret the coefficients and findings.
(c) Construct a table that compares the two models as to R-squares, assumptions, white noise, etc. Which model is best here and why?

 

year 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 jan 47 218 16 5 3 0 2 32 47 21 88 6 30 16 79 75 4 28 17 137 15 21 feb 72 24 57 31 2 18 7 10 12 54 158 37 1 63 57 39 42 3 103 69 48 26 mar 73 56 103 34 46 43 50 62 158 180 131 117 33 75 154 18 20 10 105 192 55 107 apr 182 182 136 143 119 162 125 132 258 169 191 234 139 758 206 86 129 171 142 214 130 272 may 314 318 241 240 208 573 512 122 143 255 476 201 304 326 121 267 130 381 217 291 170 506 jun 383 289 136 248 97 292 270 316 119 129 296 274 324 160 111 124 287 184 86 146 155 172 jul 82 100 148 120 68 169 126 137 72 69 103 125 146 103 37 72 85 115 108 81 92 99 aug 63 79 53 69 86 44 179 123 80 74 103 60 55 57 38 45 33 45 90 119 81 73 sep 104 54 47 86 65 32 305 133 88 52 110 8 57 51 38 21 42 17 39 51 108 87.01 oct 86 17 64 121 57 26 78 18 76 86 21 66 108 23 37 62 73 40 123 66.01 nov 28 9 50 112 103 55 154 153 44 7 15 3 B 7 50 43 83 19.01 dec 6 15 26 22 101 1 28 25 42 19 45 50 132 15 53 18 20 83 19 12 66 72.01 total 1440 1361 1077 1231 955 1415 1836 1263 1139 1115 1737 1181 1382 1691 936 907 888 1177 976 1430 1126 1520.01