Table 2.

Decomposition and time series regression with external variables and interventions for Georgia, USA, divisions 1–9 and combinations 1–3, 4–6, and 7–9

Division	Univariate regression		Decomposition				Custom time series regression
	PCP (P < 0.05)	TMP (R 2)	R 2	Holdout	Norm	White noise	Variables	R 2	Holdout	Norm	White noise
Georgia	Yes	0.51	0.73	0.66	Yes	No, close	T, PCP, TMP (lag 6)	0.64	0.66	Yes	No
1	No	0.26	0.44	0.13	No, close	Yes	TMP (lag 6) IV: July 2007	0.40	0.02	Yes	Yes
2	Yes	0.30	0.55	0.40	Yes	No	T, PCP, TMP	0.46 Robusta	0.48	Yes	Yes
3	Yes	0.12	0.27	–0.37	Yes	No	PCP, TMP IV: June 2005	0.36	0.03	Yes	Yes
4	No	0.17	0.37	0.19	No, close	Yes	T, TMP (lag 6) IV: Jan 2007	0.42	0.26	Yes	Yes
5	No	0.19	0.41	0.09	Yes	Yes	TMP (lag 6)	0.33 Robust	0.31	No, close	Yes
6	Yes	0.10	0.47	0.21	Yes	No	PCP, TMP (lag 6) IV: Aug 2001 IV: June 2003	0.47	0.45	Yes	Yes
7	No	0.13	0.31	0.24	No, close	Yes	TMP IV: Aug 2001 IV: Aug 2003	0.35	0.20	Yes	Yes
8	No	0.15	0.29	0.11	Yes	Yes	TMP IV: Aug 2002	0.27	0.00	No	Yes
9	Yes	0.25	0.36	0.12	Yes	No	PCP (lag 6), TMP	0.31	0.09	No, close	Yes
1, 2, and 3	Yes	0.33	0.49	0.43	Yes	No	PCP, TMP (lag 6) IV: June 2005	0.52	0.46	Yes	Yes
4, 5, and 6	Yes	0.27	0.55	0.43	Yes	No	TMP (lag 6) IV: May 2000 IV: June 2003	0.57	0.34	No, close	Yes
7, 8, and 9	Yes	0.36	0.51	0.22	Yes	No	TMP	0.36	0.28	Yes	Yes

T trend, PCP precipitation, TMP temperature, IV intervention

^aRobust method improved normality and model results