ࡱ >
_ bjbjJJ (3b(3b * $ P t N, :
+ + + + + + + $ 4. 0 n + E
@
+ ,
R #% z
+ : g# , # Mio _ #
% , 0 N, # R X1 _ X1 # X1 #
+ +
N,
X1
L b : Supplementary Material
Table of contents
PageEdge weights matrix of estimated network (Table S1)2Unstandardized values of centrality measures (Figure S1)3 Network stability (Appendix A1)4Stability of centrality indices (Figure S2)4Bootstrapped 95% confidence intervals of estimated edge weights (Figure S3)5Bootstrapped difference tests between non-zero edge weights (Figure S4)6Bootstrapped difference tests on node strength (Figure S5)7Bootstrapped difference tests on node closeness (Figure S6)8Bootstrapped difference tests on node betweenness (Figure S7)9References10
Table S1: Edge weights matrix of the GDS items, ALDS score and dementia follow-up
up.
AbbreviationMeaningGDS1UnsatisfiedGDS2Dropping activitiesGDS3Life emptyGDS4Often boredGDS5Bad spiritsGDS6AfraidGDS7UnhappyGDS8HelplessGDS9Staying at homeGDS10Memory problemsGDS11Awful being aliveGDS12WorthlessGDS13Lack of energyGDS14HopelessGDS15Others better offDEMDementia at follow-upALDSDecreased functional ability
*Numbers reflect the edge weights between two nodes. Blanks indicate two nodes are not connected in the sparse network. Abbreviations: GDS = Geriatric Depression Scale; ALDS = Decreased functional ability measured using the Academic Medical Center Linear Disability Scale; DEM = Dementia at follow-up
Figure S1 : Unstandardized values of centrality measures.
Appendix A1 on methods: Network stability
To assess the stability of the network we used two bootstrapping methods implemented in the R package bootnet ADDIN EN.CITE Epskamp20176216(Epskamp, 2017; Epskamp and Fried, 2017)6216621617Epskamp, SachaBorsboom, DennyFried, Eiko I.Estimating psychological networks and their accuracy: A tutorial paperBehavior Research MethodsBehavior Research Methods2017March 241554-3528journal articlehttps://doi.org/10.3758/s13428-017-0862-1https://link.springer.com/content/pdf/10.3758%2Fs13428-017-0862-1.pdf10.3758/s13428-017-0862-1Epskamp201759215921592109Epskamp, S.Fried, E.Bootstrap Methods for Various Network Estimation Routines2017(Epskamp et al., 2017; Epskamp and Fried, 2017). At first a case-dropping subset bootstrap was used, in which 1000 subsets of rows were used to estimate the network and its parameters. With this method the stability of centrality indices was checked, which were then quantified by calculating correlation stability coefficients between centrality indices obtained for the full network with those obtained for networks of subsets of the data.
Strength and closeness reached high values, meaning the direct and indirect connectivity in the network can be replicated well by running the analysis over samples of the study population (Figure S2). Betweenness has a lower correlation stability, which implies it is less certain which items are most important in connecting other nodes with each other. Correlation stability for strength was 0.75, which means that 75% of the cases could be dropped to retain a correlation of at least 0.7 with the original strength parameter when applying a significance level of 0.05. For closeness this correlation stability was 0.44, and for betweenness 0.28.
Figure S2: Stability of centrality indices of the estimated network
Secondly, we used a non-parametric bootstrap, in which data rows are resampled with replacement, to investigate the accuracy of the network by assessing sampling variability in edge-weights (Figure S3). Subsequently, this non-parametric bootstrap was used to test (1) which edge-weights differ significantly from each other (Results in Figure S4), and secondly, to test differences in nodes centrality indices for significance (Results in Figure S5-7). For further details on these bootstrapping methods we refer to literature ADDIN EN.CITE Epskamp20176216(Epskamp, 2017)6216621617Epskamp, SachaBorsboom, DennyFried, Eiko I.Estimating psychological networks and their accuracy: A tutorial paperBehavior Research MethodsBehavior Research Methods2017March 241554-3528journal articlehttps://doi.org/10.3758/s13428-017-0862-1https://link.springer.com/content/pdf/10.3758%2Fs13428-017-0862-1.pdf10.3758/s13428-017-0862-1(Epskamp et al., 2017).
Figure S3: Bootstrapped 95% confidence intervals of estimated edge weights.
Figure S 4 : B o o t s t r a p p e d d i f f e r e n c e t e s t s b e t w e e n n o n - z e r o e d g e w e i g h t s ( = . 0 5 ) B l a c k b o x e s i n d i c a t e s i g n i f i c a n t d i f f e r e n c e s i n e d g e w e i g h t s .
F i g u r e S 5 : B o o t s t r a p p e d d i f f e r e n c e t e s t s o n n o d e s t r e n g t h ( = . 0 5 ) B l a c k b o x e s i n d i c a t e s i g n i f i c a n t d i f f e r e n c e s w i t h r e g a r d t o s t r e n g h o f n o d e s ; o n t h e d i a g o n a l t h e u n s t a n d a r d i z e d s t r e n g t h v a l u e s o f t h e i n d i v i d u a l n o d e s a r e s h o w n .
F i g u r e S 6 : B o o t s t r a p p e d d i f f e r e n c e t e s t s o n n o d e c l o s e n e s s ( = . 0 5 ) B l a c k b o x e s i n d i c a t e s i g n i f i c a n t d i f f e r e n c e s w i t h r e g a r d t o + 0 1 < [ ` b e g h w 7 8 > ? @ A B ű}Ţ}}űlXű}ű}} 'hZ! h\; 5CJ OJ QJ aJ mH sH !hA' 5CJ OJ QJ aJ mH sH !hZ! 5CJ OJ QJ aJ mH sH 'hZ! hb] 5CJ OJ QJ aJ mH sH hZ! hHm CJ OJ QJ aJ 'hZ! hHm 5CJ OJ QJ aJ mH sH $hZ! hHm CJ OJ QJ aJ mH sH $hZ! hHm CJ OJ QJ aJ mH sH 'hZ! hZ! 5CJ OJ QJ aJ mH sH " ) * + 0 1 e g h _ $If gdZ! kd $$If F 0 9!% !
t 0 %6 4 4
l a p ytb] $If gdDR gdHm g h v m d $If gdDR $If gdZ! kd $$If F 0 9!% !
t 0 %6 4 4
l a p ytb] v m d $If gdDR $If gd\; kd> $$If F 0 9!% !
t 0 %6 4 4
l a p ytb] v i ` $If gdDR
&