Table 1.

Summary of statistical method and result for each comparison. (Habitats are abbreviated according to their relative elevation; high refers to cloud forest (2100 m), mid- to lower montane rainforest (1000 m) and low to lowland rainforest (400 m).)

species	treatments compared	comparison	approach (one-sided unless stated)	no. observations per habitat high, mid, low/closed, open	test statistic	model		direction
						test statistic	p-value
survival results
A. baeza	high, mid and low	nest extinction	logrank (Mantel-Cox) test	H: 14, M: 18, L: 15	Χ2	41.37	<0.0001	as predicted: high>mid>low
A. baeza	high and mid	proportion of spiders remaining	HLM (mixed)	H: 12, M: 9	F	11.74	0.007	as predicted: high>mid
dispersal results
A. baeza	high and mid	number of propagules	HLM (mixed)	H: 18, M: 17	F	7.96	0.009	as predicted: high>mid
A. baeza	high and mid	propagule survival	logrank (Mantel–Cox) test	H: 47, M: 18	Χ2	2.88	0.045	as predicted: high>mid
rain exclosure
A. baeza	low, exposed versus covered	proportion of spiders remaining	HLM (GLIMMIX)	C: 22, O: 22	F	3.68	0.032	as predicted: covered>exposed
A. baeza	low, exposed versus covered	number of web strands on plant	HLM (mixed)	C: 17, O: 19	F	5.58	0.05	as predicted: covered>exposed
A. eximius	low, exposed versus covered	proportion of spiders remaining	Wilcoxon test	C: 11, O: 10	Χ2	0.24	0.31	as predicted: covered>exposed
A. eximius	low, exposed versus covered	nest volume	t-test	C: 11, O: 11	T	1.85	0.04	as predicted: covered>exposed
predation
A. baeza	high, mid and low	presence/absence of ants	two-sided nested model	H: 18, M: 17, L: 50	Χ2	26	0.038	as predicted: high<mid<low
A. baeza	high, mid and low	presence/absence of spiders	two-sided nested model	H: 18, M: 17, L: 50	Χ2	42.9	0.0002	opposite of predicted: high>mid>low