RSiena Workshop - Bipartite
Robert Krause, Christian Steglich & Tom Snijders
22 juni 2018
This R-script guides you through an RSiena analysis for bi-partite network evolution. The data for this tutorial can be found on the RSiena website.
(1) Preparatory steps
First we need to download and extract the data. Unfortuantely we do this outside this script.
If not done yet, install the necessary packages.
install.packages("RSiena", repos = "http://cran.us.r-project.org")
install.packages("network", repos = "http://cran.us.r-project.org")
install.packages("sna", repos = "http://cran.us.r-project.org")To be sure, let us check the package version of RSiena. It should be 1.2-9 or higher.
packageVersion("RSiena")## [1] '1.2.12'Let us load the packages so that we can use them.
library(RSiena)
library(sna)
library(network)Now we can read in the data:
Friendship:
friendship1 <- as.matrix(read.table("Glasgow-1-net.dat"))
friendship2 <- as.matrix(read.table("Glasgow-2-net.dat"))Demographics:
demographics <- as.matrix(read.table("Glasgow-sex-birthyear.dat"))Bipartite
leisure1 <- as.matrix(read.table("GL-1-lsr.dat"))
leisure2 <- as.matrix(read.table("GL-2-lsr.dat"))
music1 <- as.matrix(read.table("GL-1-mus.dat"))
music2 <- as.matrix(read.table("GL-2-mus.dat"))
drugs1 <- as.matrix(read.table("GL-1-drg.dat"))
drugs2 <- as.matrix(read.table("GL-2-drg.dat"))To keep the example simple we will recode the data. All the links in our data are valued. First we make the affiliations binary:
leisure1[leisure1 %in% c(2,3,4)] <- 0
leisure2[leisure2 %in% c(2,3,4)] <- 0
drugs1[drugs1 %in% c(1)] <- 0
drugs1[drugs1 %in% c(2,3,4)] <- 1
drugs2[drugs2 %in% c(1)] <- 0
drugs2[drugs2 %in% c(2,3,4)] <- 1The friendship data has also be reworked. It is valued and missing data is not coded correctly.
friendship1[friendship1 == 2] <- 1
friendship1[friendship1 == 9] <- NA
friendship2[friendship2 == 2] <- 1
friendship2[friendship2 == 9] <- NAWe do not recode the demographic variables. They only contain 1 missing value on the birht year variable. This variable will not be used in this script.
(2) Inspection of the data
Let’s move on. What is the dimensions of our data set?
nrpupils <- dim(leisure1)[1]
nrleisureItems <- dim(leisure1)[2]
nrmusicItems <- dim(music1)[2]
nrdrugsItems <- dim(drugs1)[2]Let’s now look at some descriptives:
colMeans(leisure1)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.84375 0.21875 0.49375 0.10625 0.44375 0.43750 0.43750 0.26875 0.08750
## V10 V11 V12 V13 V14 V15
## 0.03125 0.03125 0.00625 0.46875 0.04375 0.05625colMeans(leisure2)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.79375 0.10625 0.36250 0.03125 0.45625 0.50000 0.28125 0.20625 0.08125
## V10 V11 V12 V13 V14 V15
## 0.01875 0.00625 0.00625 0.43750 0.01875 0.06250colMeans(music1)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.30000 0.65000 0.20625 0.56250 0.12500 0.54375 0.01875 0.59375 0.12500
## V10 V11 V12 V13 V14 V15 V16
## 0.03125 0.03125 0.09375 0.06875 0.14375 0.26875 0.05625colMeans(music2)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.26250 0.64375 0.10625 0.50625 0.06250 0.41250 0.03750 0.50000 0.24375
## V10 V11 V12 V13 V14 V15 V16
## 0.04375 0.02500 0.11875 0.06250 0.06250 0.18750 0.05000colMeans(drugs1)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.27500 0.06250 0.03125 0.06250 0.06250 0.03750 0.01250 0.01250 0.01250
## V10 V11 V12 V13 V14
## 0.00625 0.01875 0.08125 0.02500 0.00000colMeans(drugs2)## V1 V2 V3 V4 V5 V6 V7 V8 V9
## 0.31875 0.03750 0.04375 0.11875 0.06250 0.08125 0.00625 0.01250 0.01250
## V10 V11 V12 V13 V14
## 0.00000 0.02500 0.05000 0.01875 0.00625In previos scripts we discussed how to inspect change over time. We will skip this for now.
(3) Preparing variables for RSiena
First we define the different nodesets:
pupils <- sienaNodeSet(nrpupils, nodeSetName = "pupils")
leisureItems <- sienaNodeSet(nrleisureItems, nodeSetName = "leisureItems")
musicItems <- sienaNodeSet(nrmusicItems, nodeSetName = "musicItems")
drugsItems <- sienaNodeSet(nrdrugsItems, nodeSetName = "drugsItems")Now we identify the dependent variables.
Bipartite network:
leisure <- sienaDependent(array(c(leisure1, leisure2),
dim = c(nrpupils, nrleisureItems,2)),
"bipartite", nodeSet = c("pupils", "leisureItems"))
music <- sienaDependent(array(c(music1, music2),
dim = c(nrpupils, nrmusicItems,2)),
"bipartite", nodeSet = c("pupils", "musicItems"))
drugs <- sienaDependent(array(c(drugs1, drugs2),
dim = c(nrpupils, nrdrugsItems,2)),
"bipartite", nodeSet = c("pupils","drugsItems"))One-mode network:
friendship <- sienaDependent(array(c(friendship1, friendship2),
dim = c(nrpupils, nrpupils,2)), nodeSet = "pupils")Now we include gender.
sex.F <- coCovar(demographics[, 1], nodeSet = "pupils")All the data is ready now, so let’s combine it:
bipData <- sienaDataCreate(friendship, leisure, music, drugs, sex.F,
nodeSets = list(pupils, leisureItems, musicItems, drugsItems))(4)Specifying the model
The analysis of all this considerable data set can be time-consuming. Of course there is also the possibility to use less than 3 (e.g., 1) of the 2-mode networks.
First we get the effects table for model specification:
bipEffects <- getEffects(bipData)We can generate an initial descriptive outputfile:
print01Report(bipData,modelname = 'bipEffects-descriptive')File “netDynamics-descriptive.out” was created in the working directory. You can use getwd() to find your current directory. But I recommend to use Rprojects in RStudio.
Take a look at the generated output file “Bipartite-illustration.out” to see how RSiena interpreted the directives provided and to get an impression of the data (eyeball the degrees, look at occurrence of missings, look at Jaccard indices).
Let’s add some effects:
bipEffects <- includeEffects(bipEffects, inPopSqrt,outActSqrt,gwespFF, name = "friendship")## effectName include fix test
## 1 friendship: GWESP I -> K -> J (#) TRUE FALSE FALSE
## 2 friendship: indegree - popularity (sqrt) TRUE FALSE FALSE
## 3 friendship: outdegree - activity (sqrt) TRUE FALSE FALSE
## initialValue parm
## 1 0 69
## 2 0 0
## 3 0 1bipEffects <- includeEffects(bipEffects, from, name = "friendship",
interaction1 = "leisure")## effectName include fix test initialValue parm
## 1 friendship: from leisure agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, from, name = "friendship",
interaction1 = "music")## effectName include fix test initialValue parm
## 1 friendship: from music agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, from, name = "friendship",
interaction1 = "drugs")## effectName include fix test initialValue parm
## 1 friendship: from drugs agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, cycle4, name = "leisure")## effectName include fix test initialValue parm
## 1 leisure: 4-cycles (#) TRUE FALSE FALSE 0 1bipEffects <- includeEffects(bipEffects, cycle4, name = "music")## effectName include fix test initialValue parm
## 1 music: 4-cycles (#) TRUE FALSE FALSE 0 1bipEffects <- includeEffects(bipEffects, cycle4, name = "drugs")## effectName include fix test initialValue parm
## 1 drugs: 4-cycles (#) TRUE FALSE FALSE 0 1bipEffects <- includeEffects(bipEffects, to, name = "leisure",
interaction1 = "friendship")## effectName include fix test initialValue parm
## 1 leisure: friendship to agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, to, name = "music",
interaction1 = "friendship")## effectName include fix test initialValue parm
## 1 music: friendship to agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, to, name = "drugs",
interaction1 = "friendship")## effectName include fix test initialValue parm
## 1 drugs: friendship to agreement TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, egoX,altX,simX, name = "friendship",
interaction1 = "sex.F")## effectName include fix test initialValue parm
## 1 friendship: sex.F alter TRUE FALSE FALSE 0 0
## 2 friendship: sex.F ego TRUE FALSE FALSE 0 0
## 3 friendship: sex.F similarity TRUE FALSE FALSE 0 0bipEffects(5) Model estimation
First we specify our algorithm:
bipModel <- sienaAlgorithmCreate(projname = 'bipartite-Glasgow-results',
seed = 432155)Finally we can start with the estimation:
(bipResults <- siena07(bipModel, data = bipData, effects = bipEffects, useCluster = TRUE,
nbrNodes = 4))## Estimates, standard errors and convergence t-ratios
##
## Estimate Standard Convergence
## Error t-ratio
## 1. rate basic rate parameter friendship 14.0728 ( 1.2619 ) -0.0610
## 2. eval friendship: outdegree (density) -0.7072 ( 0.3788 ) 0.0605
## 3. eval friendship: reciprocity 1.8380 ( 0.1262 ) 0.0649
## 4. eval friendship: GWESP I -> K -> J (69) 1.6384 ( 0.0969 ) 0.0557
## 5. eval friendship: indegree - popularity (sqrt) -0.2137 ( 0.1049 ) 0.0389
## 6. eval friendship: outdegree - activity (sqrt) -0.7077 ( 0.1230 ) 0.0462
## 7. eval friendship: sex.F alter -0.1373 ( 0.1270 ) 0.0468
## 8. eval friendship: sex.F ego 0.1094 ( 0.1405 ) 0.0769
## 9. eval friendship: sex.F similarity 0.7667 ( 0.1150 ) 0.0675
## 10. eval friendship: from leisure agreement 0.0542 ( 0.0492 ) 0.0351
## 11. eval friendship: from music agreement 0.0513 ( 0.0442 ) 0.0349
## 12. eval friendship: from drugs agreement -0.2148 ( 0.1117 ) -0.0005
## 13. rate basic rate parameter leisure 4.4671 ( 0.3621 ) -0.0489
## 14. eval leisure: outdegree (density) -1.8449 ( 0.1112 ) 0.0376
## 15. eval leisure: 4-cycles (1) 0.0071 ( 0.0011 ) 0.0294
## 16. eval leisure: friendship to agreement 0.3557 ( 0.0688 ) 0.0431
## 17. rate basic rate parameter music 5.1645 ( 0.3910 ) -0.0178
## 18. eval music: outdegree (density) -1.7567 ( 0.0903 ) -0.0886
## 19. eval music: 4-cycles (1) 0.0075 ( 0.0011 ) -0.1047
## 20. eval music: friendship to agreement 0.3198 ( 0.0586 ) -0.0847
## 21. rate basic rate parameter drugs 2.4501 ( 0.4181 ) -0.1695
## 22. eval drugs: outdegree (density) -3.5416 ( 0.3027 ) -0.1491
## 23. eval drugs: 4-cycles (1) 0.1617 ( 0.0330 ) -0.1842
## 24. eval drugs: friendship to agreement 1.2208 ( 0.2376 ) -0.1487
##
## Overall maximum convergence ratio: 0.3061
##
##
## Total of 2032 iteration steps.The convergence is nearly there. But let’s be save and start again. Maybe it also helps to increase phase 3 simulations to get better estiamtes of the standard errors.
bipModel <- sienaAlgorithmCreate(projname = 'bipartite-Glasgow-results',
seed = 432155, n3 = 2000)
bipResults <- siena07(bipModel, data = bipData, effects = bipEffects,
prevAns = bipResults, returnDeps = TRUE, useCluster = TRUE,
nbrNodes = 4)
bipResults## Estimates, standard errors and convergence t-ratios
##
## Estimate Standard Convergence
## Error t-ratio
## 1. rate basic rate parameter friendship 14.1252 ( 1.4351 ) 0.0151
## 2. eval friendship: outdegree (density) -0.7407 ( 0.3667 ) -0.0206
## 3. eval friendship: reciprocity 1.8406 ( 0.1136 ) -0.0146
## 4. eval friendship: GWESP I -> K -> J (69) 1.6344 ( 0.0996 ) -0.0162
## 5. eval friendship: indegree - popularity (sqrt) -0.2087 ( 0.1024 ) -0.0284
## 6. eval friendship: outdegree - activity (sqrt) -0.6979 ( 0.1138 ) -0.0275
## 7. eval friendship: sex.F alter -0.1302 ( 0.1178 ) 0.0499
## 8. eval friendship: sex.F ego 0.1027 ( 0.1535 ) 0.0386
## 9. eval friendship: sex.F similarity 0.7605 ( 0.1171 ) -0.0182
## 10. eval friendship: from leisure agreement 0.0537 ( 0.0534 ) -0.0483
## 11. eval friendship: from music agreement 0.0501 ( 0.0467 ) -0.0195
## 12. eval friendship: from drugs agreement -0.2125 ( 0.1308 ) -0.0206
## 13. rate basic rate parameter leisure 4.4715 ( 0.3451 ) -0.0099
## 14. eval leisure: outdegree (density) -1.8495 ( 0.1041 ) 0.0186
## 15. eval leisure: 4-cycles (1) 0.0072 ( 0.0012 ) 0.0304
## 16. eval leisure: friendship to agreement 0.3525 ( 0.0659 ) 0.0252
## 17. rate basic rate parameter music 5.1511 ( 0.4072 ) -0.0272
## 18. eval music: outdegree (density) -1.7612 ( 0.0905 ) -0.0233
## 19. eval music: 4-cycles (1) 0.0076 ( 0.0011 ) -0.0217
## 20. eval music: friendship to agreement 0.3197 ( 0.0612 ) -0.0073
## 21. rate basic rate parameter drugs 2.4686 ( 0.3898 ) 0.0482
## 22. eval drugs: outdegree (density) -3.5485 ( 0.3228 ) 0.0353
## 23. eval drugs: 4-cycles (1) 0.1665 ( 0.0356 ) 0.0327
## 24. eval drugs: friendship to agreement 1.2102 ( 0.2180 ) 0.0132
##
## Overall maximum convergence ratio: 0.1604
##
##
## Total of 3220 iteration steps.It takes quite long to estimate. So let’s better save the results.
save.image("bipResults1.RData")(6) Goodness of fit
Now let us assess goodness of fit for the three estimated models. This is done separately for the various dependent variables as indicated by the varName parameter in sienaGOF.
Friendship
Outdegree
gofo1f <- sienaGOF(bipResults, OutdegreeDistribution, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gofo1f)
That does not fit well.
Indegree
gofi1f <- sienaGOF(bipResults, IndegreeDistribution, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gofi1f)
This fits very well.
Triad Census
gof1.triads <- sienaGOF(bipResults,TriadCensus, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gof1.triads,center = TRUE,scale = TRUE)
This fits quite badly!
Especially 030C, 120D, 120C and 300.
Leisure
Outdegree
gofo1l <- sienaGOF(bipResults, OutdegreeDistribution, join = TRUE,
varName = "leisure")## > Completed 2000 calculationsplot(gofo1l)
That, too, fits bad.
gofi1l <- sienaGOF(bipResults, IndegreeDistribution, join = TRUE,
varName = "leisure")## > Completed 2000 calculationsplot(gofi1l)
This does not fit well.
Music
Outdegree
gofo1m <- sienaGOF(bipResults, OutdegreeDistribution, join = TRUE,
varName = "music")## > Completed 2000 calculationsplot(gofo1m)
This does not fit too well.
Indegree
gofi1m <- sienaGOF(bipResults, IndegreeDistribution, join = TRUE,
varName = "music")## > Completed 2000 calculationsplot(gofi1m)
This fits ok.
Drugs
Outdegree
gofo1d <- sienaGOF(bipResults, OutdegreeDistribution, join = TRUE,
varName = "drugs")## > Completed 2000 calculationsplot(gofo1d)
This is ok.
Indegree
gofi1d <- sienaGOF(bipResults, IndegreeDistribution, join = TRUE,
varName = "drugs")## > Completed 2000 calculationsplot(gofi1d)
This fits very well.
To conclude: We need better fit for:
Friendship, Leisure and Music
Improved model
Let us include some effects for the distribution of Leisure and Music To keep the model equal on the bipartite part, we add the same effect also for Drugs.
bipEffects <- includeEffects(bipEffects, outActSqrt, name = "leisure")## effectName include fix test initialValue
## 1 leisure: outdegree - activity (sqrt) TRUE FALSE FALSE 0
## parm
## 1 0bipEffects <- includeEffects(bipEffects, outActSqrt, name = "music")## effectName include fix test initialValue parm
## 1 music: outdegree - activity (sqrt) TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, inPopSqrt, name = "leisure")## effectName include fix test initialValue
## 1 leisure: indegree - popularity (sqrt) TRUE FALSE FALSE 0
## parm
## 1 0bipEffects <- includeEffects(bipEffects, inPopSqrt, name = "music")## effectName include fix test initialValue
## 1 music: indegree - popularity (sqrt) TRUE FALSE FALSE 0
## parm
## 1 0bipEffects <- includeEffects(bipEffects, egoX, name = "leisure",
interaction1 = "sex.F")## effectName include fix test initialValue parm
## 1 leisure: sex.F ego TRUE FALSE FALSE 0 0bipEffects <- includeEffects(bipEffects, egoX, name = "music",
interaction1 = "sex.F")## effectName include fix test initialValue parm
## 1 music: sex.F ego TRUE FALSE FALSE 0 0Further we can add some effects for better network fit for friendship:
bipEffects <- includeEffects(bipEffects, gwespBB, gwespBF, reciAct, inActSqrt, name = "friendship")## effectName include fix test initialValue
## 1 friendship: GWESP I <- K <- J (#) TRUE FALSE FALSE 0
## 2 friendship: GWESP I <- K -> J (#) TRUE FALSE FALSE 0
## 3 friendship: indegree - activity (sqrt) TRUE FALSE FALSE 0
## 4 friendship: rec.degree^(1/#) - activity TRUE FALSE FALSE 0
## parm
## 1 69
## 2 69
## 3 0
## 4 1bipModel <- sienaAlgorithmCreate(projname = 'bipartite-Glasgow-results',
seed = 432155, n3 = 2000)
bipResults2 <- siena07(bipModel, data = bipData, effects = bipEffects,
prevAns = bipResults, returnDeps = TRUE,
useCluster = TRUE, nbrNodes = 2)
bipModel <- sienaAlgorithmCreate(projname = 'bipartite-Glasgow-results',
seed = 432155, n3 = 2000, nsub = 1,
n2start = 1200)
bipResults2 <- siena07(bipModel, data = bipData, effects = bipEffects,
prevAns = bipResults2, returnDeps = TRUE,
useCluster = TRUE, nbrNodes = 2)
bipResults2## Estimates, standard errors and convergence t-ratios
##
## Estimate Standard Convergence
## Error t-ratio
## 1. rate basic rate parameter friendship 13.4080 ( 4.8393 ) -0.0186
## 2. eval friendship: outdegree (density) -2.4413 ( 0.7457 ) -0.0006
## 3. eval friendship: reciprocity 3.1072 ( 0.4096 ) 0.0019
## 4. eval friendship: GWESP I -> K -> J (69) 1.9731 ( 0.9035 ) -0.0027
## 5. eval friendship: GWESP I <- K <- J (69) 0.6210 ( 0.4147 ) 0.0051
## 6. eval friendship: GWESP I <- K -> J (69) -0.8457 ( 1.2321 ) -0.0101
## 7. eval friendship: indegree - popularity (sqrt) -0.0752 ( 0.1345 ) -0.0008
## 8. eval friendship: indegree - activity (sqrt) 0.1592 ( 0.2456 ) -0.0034
## 9. eval friendship: outdegree - activity (sqrt) -0.1591 ( 0.2797 ) -0.0097
## 10. eval friendship: rec.degree^(1/1) - activity -0.2900 ( 0.1211 ) -0.0051
## 11. eval friendship: sex.F alter -0.1189 ( 0.1417 ) 0.0015
## 12. eval friendship: sex.F ego 0.0961 ( 0.1532 ) 0.0061
## 13. eval friendship: sex.F similarity 0.7055 ( 0.1378 ) -0.0187
## 14. eval friendship: from leisure agreement 0.0710 ( 0.0591 ) -0.0107
## 15. eval friendship: from music agreement 0.0567 ( 0.0494 ) 0.0071
## 16. eval friendship: from drugs agreement -0.2064 ( 0.1611 ) -0.0164
## 17. rate basic rate parameter leisure 4.7391 ( 0.6046 ) -0.0181
## 18. eval leisure: outdegree (density) -3.7676 ( 0.7570 ) 0.0250
## 19. eval leisure: 4-cycles (1) 0.0001 ( 0.0025 ) 0.0079
## 20. eval leisure: indegree - popularity (sqrt) 0.2714 ( 0.0581 ) 0.0199
## 21. eval leisure: outdegree - activity (sqrt) 0.3065 ( 0.2060 ) 0.0169
## 22. eval leisure: sex.F ego -0.0260 ( 0.0982 ) -0.0123
## 23. eval leisure: friendship to agreement 0.1890 ( 0.0735 ) 0.0251
## 24. rate basic rate parameter music 5.3070 ( 0.4139 ) 0.0018
## 25. eval music: outdegree (density) -2.3231 ( 0.5384 ) -0.0013
## 26. eval music: 4-cycles (1) 0.0051 ( 0.0023 ) -0.0068
## 27. eval music: indegree - popularity (sqrt) 0.1010 ( 0.0482 ) -0.0010
## 28. eval music: outdegree - activity (sqrt) 0.0613 ( 0.1332 ) -0.0047
## 29. eval music: sex.F ego 0.0386 ( 0.1053 ) 0.0104
## 30. eval music: friendship to agreement 0.2488 ( 0.0838 ) 0.0132
## 31. rate basic rate parameter drugs 2.4547 ( 0.4396 ) -0.0233
## 32. eval drugs: outdegree (density) -3.6168 ( 0.3926 ) -0.0027
## 33. eval drugs: 4-cycles (1) 0.1653 ( 0.0363 ) -0.0051
## 34. eval drugs: friendship to agreement 1.2743 ( 0.2657 ) 0.0040
##
## Overall maximum convergence ratio: 0.1759
##
##
## Total of 3200 iteration steps.The model is converged. Let’s check the fit again.
Goodness of fit
Now let us assess goodness of fit for the three estimated models. This is done separately for the various dependent variables as indicated by the varName parameter in sienaGOF.
Friendship
Outdegree
gofo2f <- sienaGOF(bipResults2, OutdegreeDistribution, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gofo2f)
Still not ideal.
Indegree
gofi2f <- sienaGOF(bipResults2, IndegreeDistribution, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gofi2f)
This fits very well.
Triad Census
gof2.triads <- sienaGOF(bipResults2,TriadCensus, join = TRUE,
varName = "friendship")## > Completed 2000 calculationsplot(gof2.triads,center = TRUE,scale = TRUE)
This fits fine now.
120C and 300 are not perfect, but overall it is fine overall.
Leisure
Outdegree
gofo2l <- sienaGOF(bipResults2, OutdegreeDistribution, join = TRUE,
varName = "leisure")## > Completed 2000 calculationsplot(gofo2l)
This, still, fits bad.
gofi2l <- sienaGOF(bipResults2, IndegreeDistribution, join = TRUE,
varName = "leisure")## > Completed 2000 calculationsplot(gofi2l)
This fits fine now.
Music
Outdegree
gofo2m <- sienaGOF(bipResults2, OutdegreeDistribution, join = TRUE,
varName = "music")## > Completed 2000 calculationsplot(gofo2m)
This is still not optimal.
Indegree
gofi2m <- sienaGOF(bipResults2, IndegreeDistribution, join = TRUE,
varName = "music")## > Completed 2000 calculationsplot(gofi2m)
This fits ok.
Drugs
Outdegree
gofo2d <- sienaGOF(bipResults2, OutdegreeDistribution, join = TRUE,
varName = "drugs")## > Completed 2000 calculationsplot(gofo2d)
This is ok.
Indegree
gofi2d <- sienaGOF(bipResults2, IndegreeDistribution, join = TRUE,
varName = "drugs")## > Completed 2000 calculationsplot(gofi2d)
This fits very well.
To conclude: We need better fit for:
Friendship, Leisure and Music