chunk_opts = list(echo = TRUE, fig.align = 'center', fig.height = 8, fig.width = 8)Multiple Correspondence Analysis - Faith Survey
1. Routine immunization
1.1 Data summary
summary(regVaxData) age sex geoLoc role
18-24:2552 f :3646 rural:4888 relLead:2452
25-34:3515 sex_nr: 623 urban:4783 member :3888
35-44:2242 m :5402 notPart:3331
45+ :1362
religion leadOpCVax
christian :3898 leadOpCVax_matters :4364
muslim :3771 leadOpCVax_ns :2361
hindu : 624 leadOpCVax_dntMatter:2742
other : 369 NA's : 204
traditional religions: 286
buddhist : 229
(Other) : 494
mostLeadEndorseCV vaxed influencerCVax
noLeadEndorseCVax :1331 unvaxed:2579 infCvax_relLead : 791
notSureLeadEndorseCVax:2218 vaxed :7092 infCvax_hWorker :3802
mostLeadEndorseCVax :6122 infCvax_other :1844
infCvax_peerFriend: 750
infCvax_FamMember :2484
factorCVax importanceRVax influencerRVax
fClearInfo :3109 regVaxVeryImp :5656 infRVax_famMember :2636
fLeaderTakingVax:2639 regVaxSwhatImp:2067 infRVax_hWorker :4489
fRecommendations: 894 regVaxnotImp :1230 infRVax_other :1412
fDoMyPart : 840 regVaxnotSure : 718 infRVax_peerFriend: 464
fOther : 839 infRVax_relLeader : 670
fMedia : 572
(Other) : 778
continuedRvax leadSupportRvax
continuedRVax :7352 noLeadSupportRvax : 921
notcontinuedRVax:2319 notSureLeadSupportRvax:1710
someLeadSupportRvax :1681
mostLeadSupportRvax :5359
howReligionInfRVax contact_type country
relDiscouragesRVax: 931 Religions For Peace:2292 Guatemala:1472
relDsntAffectRVax :1332 RDD / 3-2-1 :7379 Mali :1196
relNSRVax :1449 Iraq :1139
relEncouragesRVax :5959 Zambia :1069
Kenya : 957
Nigeria : 953
(Other) :2885 This dataset contains 9671 individuals and 48 variables, 7 qualitative variables are considered as illustrative.
1.2 Run the MCA
regVax.MCA<-MCA(regVaxData,ncp=8,quali.sup=c(1,2,3,4,5,16,17),graph=FALSE)1.3 Inertia distribution (auto text)
The inertia of the first dimensions shows if there are strong relationships between variables and suggests the number of dimensions that should be studied.
The first two dimensions of analyse express 17.39% of the total dataset inertia ; that means that 17.39% of the individuals (or variables) cloud total variability is explained by the plane. This is a very small percentage and the first plane represents a small part of the data variability. This value is greater than the reference value that equals 7.1%, the variability explained by this plane is thus significant (the reference value is the 0.95-quantile of the inertia percentages distribution obtained by simulating 532 data tables of equivalent size on the basis of a uniform distribution).
From these observations, it is important to also interpret the dimensions greater or equal to the third one.
An estimation of the right number of axis to interpret suggests to restrict the analysis to the description of the first 8 axis. These axis present an amount of inertia greater than those obtained by the 0.95-quantile of random distributions (42.06% against 27.69%). This observation suggests that only these axis are carrying a real information. As a consequence, the description will stand to these axis.
1.4 Description of the plane 1:2
Analyzing individuals
Figure 4.1 - Individuals factor map (MCA)
Analyzing specific individuals
regVaxData %>% tibble::rownames_to_column("rowname") %>%
filter(rowname %in% c(1683, 2175, 2908, 1562, 5027, 1839, 2539, 2377, 3128, 4961, 1376, 7069)) %>%
mutate(quadrant = c(3, 4, 1, 1, 1, 4, 1, 4, 4, 3, 3, 3), .before = 1) %>% arrange(quadrant) %>% kable()| quadrant | rowname | age | sex | geoLoc | role | religion | leadOpCVax | mostLeadEndorseCV | vaxed | influencerCVax | factorCVax | importanceRVax | influencerRVax | continuedRvax | leadSupportRvax | howReligionInfRVax | contact_type | country |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 1683 | 18-24 | m | urban | member | muslim | leadOpCVax_dntMatter | noLeadEndorseCVax | unvaxed | infCvax_FamMember | fRecommendations | regVaxSwhatImp | infRVax_famMember | notcontinuedRVax | someLeadSupportRvax | relDiscouragesRVax | RDD / 3-2-1 | Guatemala |
| 1 | 1839 | 18-24 | f | rural | member | muslim | leadOpCVax_dntMatter | noLeadEndorseCVax | unvaxed | infCvax_FamMember | fClearInfo | regVaxSwhatImp | infRVax_relLeader | notcontinuedRVax | someLeadSupportRvax | relDiscouragesRVax | RDD / 3-2-1 | Guatemala |
| 1 | 2175 | 18-24 | m | urban | relLead | christian | leadOpCVax_dntMatter | noLeadEndorseCVax | unvaxed | infCvax_FamMember | fRecommendations | regVaxSwhatImp | infRVax_relLeader | notcontinuedRVax | someLeadSupportRvax | relDiscouragesRVax | RDD / 3-2-1 | Guatemala |
| 1 | 2539 | 35-44 | f | rural | notPart | traditional religions | leadOpCVax_dntMatter | noLeadEndorseCVax | unvaxed | infCvax_FamMember | fDoMyPart | regVaxSwhatImp | infRVax_famMember | notcontinuedRVax | someLeadSupportRvax | relDiscouragesRVax | RDD / 3-2-1 | Guatemala |
| 3 | 1376 | 45+ | sex_nr | rural | notPart | traditional religions | leadOpCVax_matters | mostLeadEndorseCVax | vaxed | infCvax_hWorker | fRecommendations | regVaxnotImp | infRVax_hWorker | continuedRVax | mostLeadSupportRvax | relEncouragesRVax | RDD / 3-2-1 | Guatemala |
| 3 | 4961 | 35-44 | m | urban | relLead | muslim | leadOpCVax_matters | mostLeadEndorseCVax | vaxed | infCvax_hWorker | fLeaderTakingVax | regVaxnotImp | infRVax_peerFriend | continuedRVax | mostLeadSupportRvax | relEncouragesRVax | RDD / 3-2-1 | Mali |
| 3 | 5027 | 25-34 | m | urban | relLead | muslim | leadOpCVax_matters | mostLeadEndorseCVax | vaxed | infCvax_hWorker | fGovGuidance | regVaxnotImp | infRVax_hWorker | continuedRVax | mostLeadSupportRvax | relEncouragesRVax | Religions For Peace | Mali |
| 3 | 7069 | 45+ | m | urban | notPart | muslim | leadOpCVax_matters | mostLeadEndorseCVax | vaxed | infCvax_hWorker | fOther | regVaxVeryImp | infRVax_hWorker | continuedRVax | mostLeadSupportRvax | relEncouragesRVax | RDD / 3-2-1 | Nigeria |
| 4 | 1562 | 18-24 | f | rural | member | not religious | NA | notSureLeadEndorseCVax | unvaxed | infCvax_peerFriend | fOther | regVaxnotSure | infRVax_other | notcontinuedRVax | notSureLeadSupportRvax | relNSRVax | RDD / 3-2-1 | Guatemala |
| 4 | 2377 | 45+ | sex_nr | rural | notPart | other | leadOpCVax_ns | notSureLeadEndorseCVax | unvaxed | infCvax_other | fDoMyPart | regVaxnotSure | infRVax_other | continuedRVax | notSureLeadSupportRvax | relNSRVax | RDD / 3-2-1 | Guatemala |
| 4 | 2908 | 25-34 | sex_nr | rural | notPart | buddhist | leadOpCVax_ns | notSureLeadEndorseCVax | vaxed | infCvax_peerFriend | fOther | regVaxnotSure | infRVax_other | continuedRVax | notSureLeadSupportRvax | relNSRVax | RDD / 3-2-1 | Iraq |
| 4 | 3128 | 45+ | sex_nr | urban | notPart | muslim | leadOpCVax_ns | notSureLeadEndorseCVax | unvaxed | infCvax_other | fOther | regVaxnotSure | infRVax_hWorker | continuedRVax | notSureLeadSupportRvax | relNSRVax | RDD / 3-2-1 | Iraq |
Analyzing categories
plot.MCA(regVax.MCA, invisible= 'ind',col.quali.sup='#006400',label =c('var'))Warning: ggrepel: 10 unlabeled data points (too many overlaps). Consider
increasing max.overlaps
Figure 4.2 - Categories factor map (MCA)
Analyzing variables
plot.MCA(regVax.MCA, choix='var',col.quali.sup='#006400')Warning: ggrepel: 2 unlabeled data points (too many overlaps). Consider
increasing max.overlaps
Figure 4.3 - Variable representation (MCA)
The square correlation of variables with axes are not too high, but considering the large sample, those with a correlation closer to 0.5 are considered relevant: For dimension 1:
- leadSupportRvax, howReligionInfRVax, factorCVax, mostLeadEndorseCV, leadOpCVax, influencerCVax, influencerRVax
For dimension 2:
- leadSupportRvax, howReligionInfRVax.
Analyzing categories by variable
plot.MCA(regVax.MCA,invisible= c('ind','quali.sup'),col.var=c(1,1,1,1,1,2,2,2,3,3,4,4,4,4,4,5,5,5,5,5,5,5,5,6,6,6,6,7,7,7,7,7,8,8,9,9,9,9,10,10,10,10),col.quali.sup=c(11,11,11,11,12,12,12,13,13,14,14,14,15,15,15,15,15,15,15,15,15,15,16,16,17,17,17,17,17,17,17,17,17,17),cex=0.8,cex.main=0.8,cex.axis=0.8,label =c('var'))Warning: ggrepel: 7 unlabeled data points (too many overlaps). Consider
increasing max.overlaps
Figure 4.5 - Category factors, by variable (MCA)
The positive diagonal seems to distinguish pro-vaccine individuals (vaccination, its willingness) and individuals who believe their leaders or religious beliefs are in favour of vaccination (in the third quadrant) from those against vaccination (unvaccinated, those who did not continue routine immunization) and those who believe their religion or their leaders discourage vaccination (in the first quadrant). This diagonal show differences on influencers of vaccination (both regular and covid), having health workers close to those who are pro-vaccine, and family close to those who are against vaccination. Those believing religious leaders support vaccination are also close to those who feel influenced by their leaders taking the vaccine, whereas those who are not sure about religious support are close to those who feel most influenced by recommendations of influential people, public statements or media.
Duboious categories (those answering “not sure”, “other” or “prefer not to answer”) in several questions seem to be grouped in the fourth quadrant.
Analyzing categories and supporting variables
plot.MCA(regVax.MCA,invisible= 'ind',col.quali.sup='#006400',label =c('var','quali.sup')) %>% ggplotlyFigure 4.5 - Category factors, with supporting variable categories (MCA)
Looking at supportive variables, we find respondents from RFP, and those from Lenya, Zambia, Mali (and also Nigeria), as well as those who declared themselves as religious leaders, close to those who believe religious leaders support vaccination, in the third quadrant.
Jewish, and traditional religions seem to group in the first quadrant, close to those who believe religious leaders discourage vaccination.
1.5 Clustering
regVax.MCA$eig #%>% head() eigenvalue percentage of variance cumulative percentage of variance
dim 1 0.36619820 11.812845 11.81285
dim 2 0.17289135 5.577140 17.38999
dim 3 0.14953521 4.823716 22.21370
dim 4 0.13691707 4.416680 26.63038
dim 5 0.13137543 4.237917 30.86830
dim 6 0.12615782 4.069607 34.93791
dim 7 0.11372916 3.668683 38.60659
dim 8 0.10703568 3.452764 42.05935
dim 9 0.10288402 3.318839 45.37819
dim 10 0.10056265 3.243956 48.62215
dim 11 0.10033287 3.236544 51.85869
dim 12 0.09687487 3.124996 54.98369
dim 13 0.09263062 2.988085 57.97177
dim 14 0.09164328 2.956235 60.92801
dim 15 0.09014075 2.907766 63.83577
dim 16 0.08832636 2.849237 66.68501
dim 17 0.08713409 2.810777 69.49579
dim 18 0.08302324 2.678169 72.17396
dim 19 0.08161335 2.632689 74.80665
dim 20 0.07893656 2.546341 77.35299
dim 21 0.07745531 2.498558 79.85154
dim 22 0.07429881 2.396736 82.24828
dim 23 0.07273790 2.346384 84.59466
dim 24 0.07199724 2.322492 86.91716
dim 25 0.06883658 2.220535 89.13769
dim 26 0.06549124 2.112621 91.25031
dim 27 0.06120324 1.974298 93.22461
dim 28 0.05668930 1.828687 95.05330
dim 29 0.05417428 1.747557 96.80085
dim 30 0.05100437 1.645302 98.44616
dim 31 0.04816916 1.553844 100.00000Choosing 22 dimension to get >80% of variance
Performing clustering
# Run this in console to choose the number of clusters
#regVax.hcpc <- HCPC(regVax.MCA)
#Leave this on chunk for rendering
regVax.hcpc <- HCPC(regVax.MCA, nb.clust = 4)
This plot made with 4 clusters suggested by dendogram
# plot.HCPC(res.HCPC,choice='map',draw.tree=FALSE,title='Factor map', label = "none")
plot(regVax.hcpc,choice="map",ind.names=FALSE)
round(regVax.hcpc$call$t$inert.gain,3) %>% head()[1] 0.178 0.100 0.079 0.071 0.058 0.045round(regVax.MCA$eig[,1],3) %>% head()dim 1 dim 2 dim 3 dim 4 dim 5 dim 6
0.366 0.173 0.150 0.137 0.131 0.126 Defining the clusters
regVax.hcpc$desc.var$test p.value df
leadOpCVax 0.000000e+00 9
mostLeadEndorseCV 0.000000e+00 6
influencerCVax 0.000000e+00 12
factorCVax 0.000000e+00 21
importanceRVax 0.000000e+00 9
influencerRVax 0.000000e+00 12
leadSupportRvax 0.000000e+00 9
howReligionInfRVax 0.000000e+00 9
country 0.000000e+00 27
vaxed 5.549269e-311 3
continuedRvax 6.781057e-310 3
religion 1.974150e-229 27
role 2.782092e-198 6
sex 2.305549e-167 6
contact_type 2.867685e-134 3
age 1.034460e-17 9
geoLoc 5.436906e-05 3regVax.hcpc$desc.var$category %>% lapply(head)$`1`
Cla/Mod Mod/Cla Global p.value
howReligionInfRVax=relEncouragesRVax 61.97349 90.44820 61.61721 0.000000e+00
leadSupportRvax=mostLeadSupportRvax 65.42265 85.86823 55.41309 0.000000e+00
influencerCVax=infCvax_hWorker 66.54392 61.96424 39.31341 0.000000e+00
mostLeadEndorseCV=mostLeadEndorseCVax 60.07841 90.08082 63.30266 0.000000e+00
leadOpCVax=leadOpCVax_matters 68.21723 72.91207 45.12460 0.000000e+00
factorCVax=fLeaderTakingVax 71.84540 46.43644 27.28777 5.089468e-288
v.test
howReligionInfRVax=relEncouragesRVax Inf
leadSupportRvax=mostLeadSupportRvax Inf
influencerCVax=infCvax_hWorker Inf
mostLeadEndorseCV=mostLeadEndorseCVax Inf
leadOpCVax=leadOpCVax_matters Inf
factorCVax=fLeaderTakingVax 36.26844
$`2`
Cla/Mod Mod/Cla Global p.value
influencerRVax=infRVax_relLeader 82.83582 53.62319 6.927929 0.000000e+00
influencerCVax=infCvax_relLead 87.98989 67.24638 8.179092 0.000000e+00
role=relLead 18.39315 43.57488 25.354152 7.562078e-42
contact_type=Religions For Peace 18.71728 41.44928 23.699721 3.431460e-41
leadOpCVax=leadOpCVax_matters 14.84876 62.60870 45.124599 6.871624e-33
country=Nigeria 22.98006 21.15942 9.854203 4.037931e-31
v.test
influencerRVax=infRVax_relLeader Inf
influencerCVax=infCvax_relLead Inf
role=relLead 13.55342
contact_type=Religions For Peace 13.44198
leadOpCVax=leadOpCVax_matters 11.94528
country=Nigeria 11.60174
$`3`
Cla/Mod Mod/Cla Global p.value
leadOpCVax=leadOpCVax_dntMatter 59.40919 57.68414 28.352807 0.000000e+00
howReligionInfRVax=relDsntAffectRVax 71.62162 33.78187 13.773136 3.560404e-266
mostLeadEndorseCV=noLeadEndorseCVax 65.74005 30.98442 13.762796 6.468841e-198
leadSupportRvax=someLeadSupportRvax 55.68114 33.14448 17.381863 4.111474e-140
leadSupportRvax=noLeadSupportRvax 59.82628 19.51133 9.523317 2.077776e-92
influencerRVax=infRVax_famMember 41.72989 38.95184 27.256747 1.649112e-59
v.test
leadOpCVax=leadOpCVax_dntMatter Inf
howReligionInfRVax=relDsntAffectRVax 34.85513
mostLeadEndorseCV=noLeadEndorseCVax 30.01387
leadSupportRvax=someLeadSupportRvax 25.19897
leadSupportRvax=noLeadSupportRvax 20.38938
influencerRVax=infRVax_famMember 16.26859
$`4`
Cla/Mod Mod/Cla Global p.value
howReligionInfRVax=relNSRVax 66.18357 55.46559 14.982939 0
leadSupportRvax=notSureLeadSupportRvax 62.28070 61.59630 17.681729 0
influencerRVax=infRVax_other 60.19830 49.16136 14.600352 0
importanceRVax=regVaxnotSure 80.64067 33.48757 7.424258 0
factorCVax=fOther 75.80453 36.78427 8.675421 0
influencerCVax=infCvax_other 56.83297 60.61307 19.067315 0
v.test
howReligionInfRVax=relNSRVax Inf
leadSupportRvax=notSureLeadSupportRvax Inf
influencerRVax=infRVax_other Inf
importanceRVax=regVaxnotSure Inf
factorCVax=fOther Inf
influencerCVax=infCvax_other InfCluster 1 seems to group:
- Those whose leader's opinion matters to them and who believe in their endosement or support for vaccines
- Those who feel most influenced by health workers
- Those who believe the most important factor for decisions on COVID vaccination is seeing their religious leader taking the vaccineCluster 2 seems to group:
- Those who declare religious leaders as their main influencers on vaccinations
- Religious leaders or RFP membersCluster 3 seems to group:
- Those whose leader's opinion does not matter to them
- Those who believe leaders do not support vaccination, or that religious beliefs do not affect vaccination decisions
- Those who believe the most important factor for decisions on COVID vaccination is government guidanceCluster 4 seems to group:
- Those not sure about religious leaders or religious beliefs stance/influence on vaccinationImportant vars are: p.value df religion 0.000000e+00 45 Only relevant in the last clusters leadOpCVax 0.000000e+00 15 Cluster 1 has those whose leader’s opinion matters to them, Cluster 5 those who are NAs influencerCVax 0.000000e+00 20 Cluster 1 has those that feel most influenced by health workers, Cluster 2 has respondents influenced by peers and family, Cluster 6 has those influenced by other people factorCVax 0.000000e+00 35 Cluster 1 has those who believe the most important factor for decisions on COVID vaccination is seeing their religious leader taking the vaccine, for those in cluster 2 is recommendations and media, those in cluster 3 are more likely to regocnize government guidance as a factor. Those in cluster 4, in public statements by influencing people, and those in cluster 6 are likeley to believe it is because they want to do their part on vaccination influencerRVax 0.000000e+00 20 Similar to influencerCVax country 0.000000e+00 45
Cluster 1 factor leader taking vax and clear info $1 Cla/Mod Mod/Cla factorCVax=fLeaderTakingVax 86.8889731 48.20264873 factorCVax=fClearInfo 75.4261821 49.29577465 influencerCVax=infCvax_hWorker 69.0426092 55.18183729 howReligionInfRVax=relEncouragesRVax 60.9666051 76.37166281 leadOpCVax=leadOpCVax_matters 65.6049496 60.18499054 mostLeadEndorseCV=mostLeadEndorseCVax 59.8007187 76.96026908 leadSupportRvax=mostLeadSupportRvax 61.2054488 68.95101955 influencerRVax=infRVax_hWorker 63.0875473 59.53331932
$2 factor recomendations and media Cla/Mod Mod/Cla factorCVax=fRecommendations 97.0917226 41.83132530 factorCVax=fMedia 98.9510490 27.27710843 influencerCVax=infCvax_peerFriend 86.8000000 31.37349398 influencerRVax=infRVax_peerFriend 86.6379310 19.37349398 leadOpCVax=leadOpCVax_ns 28.5048708 32.43373494 leadSupportRvax=notSureLeadSupportRvax 29.3567251 24.19277108 mostLeadEndorseCV=notSureLeadEndorseCVax 27.9080252 29.83132530 howReligionInfRVax=relNSRVax 28.7094548 20.04819277 role=notPart 25.5478835 41.01204819 contact_type=RDD / 3-2-1 22.9841442 81.73493976
$3 factor government guidance Cla/Mod Mod/Cla factorCVax=fGovGuidance 98.544699 100.000000 leadSupportRvax=notSureLeadSupportRvax 9.415205 33.966245 leadOpCVax=leadOpCVax_ns 8.428632 41.983122 mostLeadEndorseCV=notSureLeadEndorseCVax 8.295762 38.818565 country=Egypt 10.911271 19.198312 role=notPart 7.205044 50.632911 howReligionInfRVax=relNSRVax 8.902692 27.215190 contact_type=RDD / 3-2-1 5.610516 87.341772 country=Pakistan 9.020902 17.299578 influencerCVax=infCvax_other 7.375271 28.691983 influencerRVax=infRVax_other 7.648725 22.784810 importanceRVax=regVaxVeryImp 5.657709 67.510549
$4 factor public statements Cla/Mod Mod/Cla factorCVax=fPubStatements 97.979798 100.000000 influencerCVax=infCvax_other 5.368764 34.020619 country=Iraq 5.882353 23.024055 vaxed=unvaxed 4.614192 40.893471 influencerRVax=infRVax_other 5.240793 25.429553 howReligionInfRVax=relNSRVax 5.037957 25.085911 leadOpCVax=leadOpCVax_ns 4.404913 35.738832 religion=buddhist 8.733624 6.872852 mostLeadEndorseCV=notSureLeadEndorseCVax 4.373309 33.333333 leadSupportRvax=notSureLeadSupportRvax 4.444444 26.116838 contact_type=RDD / 3-2-1 3.306681 83.848797 continuedRvax=notcontinuedRVax 4.010349 31.958763 mostLeadEndorseCV=noLeadEndorseCVax 4.432757 20.274914 importanceRVax=regVaxnotSure 5.013928 12.371134
$5 Cla/Mod Mod/Cla leadOpCVax=leadOpCVax.NA 100.0000000 100.0000000 religion=not religious 94.4444444 100.0000000 country=Guatemala 8.9673913 64.7058824 role=notPart 4.8934254 79.9019608 country=Brazil 9.9722992 17.6470588 factorCVax=fDoMyPart 6.4285714 26.4705882 mostLeadEndorseCV=notSureLeadEndorseCVax 4.2831380 46.5686275 sex=f 3.4009874 60.7843137 factorCVax=fOther 5.4827175 22.5490196 contact_type=RDD / 3-2-1 2.5071148 90.6862745 influencerRVax=infRVax_other 4.1784703 28.9215686 influencerCVax=infCvax_other 3.7418655 33.8235294 howReligionInfRVax=relDsntAffectRVax 3.9039039 25.4901961 howReligionInfRVax=relNSRVax 3.7267081 26.4705882 leadSupportRvax=notSureLeadSupportRvax 3.4502924 28.9215686 leadSupportRvax=someLeadSupportRvax 3.3908388 27.9411765
$6 Cla/Mod Mod/Cla factorCVax=fDoMyPart 78.2142857 35.13368984 influencerCVax=infCvax_other 57.4837310 56.68449198 influencerRVax=infRVax_other 58.5694051 44.22459893 factorCVax=fOther 66.8653159 30.00000000 leadOpCVax=leadOpCVax_ns 35.8322745 45.24064171 howReligionInfRVax=relNSRVax 40.3036577 31.22994652 leadSupportRvax=notSureLeadSupportRvax 37.8947368 34.65240642 importanceRVax=regVaxnotSure 49.8607242 19.14438503 mostLeadEndorseCV=notSureLeadEndorseCVax 34.0847610 40.42780749 country=Iraq 37.8402107 23.04812834 vaxed=unvaxed 28.9647150 39.94652406 continuedRvax=notcontinuedRVax 29.0211298 35.98930481 religion=other 47.4254743 9.35828877 sex=sex_nr 40.1284109 13.36898396 role=notPart 25.6079255 45.61497326
2. COVID immunization
2.1. Data summary
summary(covidData) age sex geoLoc role
18-24:5737 f : 7698 rural: 9051 relLead:4627
25-34:5997 sex_nr: 917 urban:10795 member :8016
35-44:4045 m :11231 notPart:7203
45+ :4067
religion leadOpCVax
christian :8287 leadOpCVax_matters :8451
muslim :7402 leadOpCVax_ns :4285
hindu :1551 leadOpCVax_dntMatter:6601
other : 592 NA's : 509
not religious : 537
traditional religions: 534
(Other) : 943
mostLeadEndorseCV vaxed influencerCVax
noLeadEndorseCVax : 2828 unvaxed: 4805 infCvax_relLead :1548
notSureLeadEndorseCVax: 4332 vaxed :15041 infCvax_hWorker :7629
mostLeadEndorseCVax :12686 infCvax_other :3930
infCvax_peerFriend:1482
infCvax_FamMember :5257
factorCVax contact_type country
fClearInfo :6972 Religions For Peace: 5586 Zambia :2125
fLeaderTakingVax:4823 RDD / 3-2-1 :14260 Kenya :2116
fRecommendations:1998 Nigeria:2087
fDoMyPart :1820 Egypt :2026
fOther :1514 Mali :2023
fMedia :1173 Nepal :2014
(Other) :1546 (Other):7455 This dataset contains 19846 individuals and 28 variables, 7 qualitative variables are considered as illustrative.
2.2 Run the MCA
dfcompleted <- missMDA::imputeMCA(covidData,ncp=14,quali.sup=c(1,2,3,4,5,11,12))
covidVax.MCA<-MCA(covidData,tab.disj = dfcompleted$tab.disj,quali.sup=c(1,2,3,4,5,11,12),graph=FALSE)2.3 Inertia distribution (autotext)
The inertia of the first dimensions shows if there are strong relationships between variables and suggests the number of dimensions that should be studied.
The first two dimensions of analyse express 21.3% of the total dataset inertia ; that means that 21.3% of the individuals (or variables) cloud total variability is explained by the plane. This is a small percentage and the first plane just represents a part of the data variability. This value is greater than the reference value that equals 13.05%, the variability explained by this plane is thus significant (the reference value is the 0.95-quantile of the inertia percentages distribution obtained by simulating 570 data tables of equivalent size on the basis of a uniform distribution).
From these observations, it is interesting to consider the next dimensions which also express a high percentage of the total inertia.
An estimation of the right number of axis to interpret suggests to restrict the analysis to the description of the first 6 axis. These axis present an amount of inertia greater than those obtained by the 0.95-quantile of random distributions (48.32% against 38.56%). This observation suggests that only these axis are carrying a real information. As a consequence, the description will stand to these axis.
2.4. Description of the plane 1:2
Analyzing individuals
Figure 2.4.1 Individuals factor map (MCA)
Horseshoe shape?
Analyzing specific individuals
regVaxData %>% tibble::rownames_to_column("rowname") %>%
filter(rowname %in% c(1683, 2175, 2908, 1562, 5027, 1839, 2539, 2377, 3128, 4961, 1376, 7069)) %>%
mutate(quadrant = c(3, 4, 1, 1, 1, 4, 1, 4, 4, 3, 3, 3), .before = 1) %>% arrange(quadrant)# A tibble: 12 × 19
quadrant rowname age sex geoLoc role religion leadOpCVax
<dbl> <chr> <fct> <fct> <fct> <fct> <fct> <fct>
1 1 1683 18-24 m urban member muslim leadOpCVa…
2 1 1839 18-24 f rural member muslim leadOpCVa…
3 1 2175 18-24 m urban relLead christian leadOpCVa…
4 1 2539 35-44 f rural notPart traditional religions leadOpCVa…
5 3 1376 45+ sex_nr rural notPart traditional religions leadOpCVa…
6 3 4961 35-44 m urban relLead muslim leadOpCVa…
7 3 5027 25-34 m urban relLead muslim leadOpCVa…
8 3 7069 45+ m urban notPart muslim leadOpCVa…
9 4 1562 18-24 f rural member not religious <NA>
10 4 2377 45+ sex_nr rural notPart other leadOpCVa…
11 4 2908 25-34 sex_nr rural notPart buddhist leadOpCVa…
12 4 3128 45+ sex_nr urban notPart muslim leadOpCVa…
# ℹ 11 more variables: mostLeadEndorseCV <fct>, vaxed <fct>,
# influencerCVax <fct>, factorCVax <fct>, importanceRVax <fct>,
# influencerRVax <fct>, continuedRvax <fct>, leadSupportRvax <fct>,
# howReligionInfRVax <fct>, contact_type <fct>, country <fct>Analyzing categories
plot.MCA(covidVax.MCA, invisible= 'ind',col.quali.sup='#006400',label =c('var'))
Figure 2.4.2 Categories factor map (MCA)
Analyzing variables
plot.MCA(covidVax.MCA, choix='var',col.quali.sup='#006400')
Figure 2.4.3 - Individuals factor map (MCA)
The square correlation of variables with axes are not too high, but considering the large sample, those with a correlation closer to 0.5 are considered relevant: For dimension 1:
- factorCVax, mostLeadEndorseCV, leadOpCVax, influencerCVax, vaxed
For dimension 2:
- factorCVax, leadOpCVax, influencerCVax
Analyzing categories by variable
plot.MCA(covidVax.MCA,invisible= c('ind','quali.sup'),col.var=c(1,1,1,2,2,2,3,3,4,4,4,4,4,5,5,5,5,5,5,5,5),col.quali.sup=c(6,6,6,6,7,7,7,8,8,9,9,9,10,10,10,10,10,10,10,10,10,10,11,11,12,12,12,12,12,12,12,12,12,12),label =c('var'))
Figure 2.4.4 - Category factors, by variable (MCA)
A horseshoe shape is observed for the influence of leader’s opinion on vaccination, as well as for the COVID vaccination endorsement of leaders and for the most important influencers of it. Those who belive in leader’s endorsement of vaccination are close to those who feel influenced by the opinion of their leaders, and those who are most influenced by them, locate on negative values of dimension 1 and positive values for dimension 2. Those who do not believe in leader’s endorsement are close to those who feel influenced by the opinion of their leaders locate close to the origin of dimension 1. Those uncertain about the influence of leaders locate at positive values of dimensions 1 and 2.
Dimension 1 seems to distinguish between those vaccinated and those who are not.
Analyzing categories and supporting variables
plot.MCA(covidVax.MCA,invisible= 'ind',col.quali.sup='#006400',label =c('var','quali.sup')) %>% ggplotlyFigure 2.4.5 - Category factors, with supporting variable categories (MCA)
When comparing to supporting variables, and as it happened with routine immunization, Nigeria, Zambia, Kenya and Mali, as well as those who declared themselves as religious leaders, seem closer to those who believe religious leaders support vaccination and to those who feel influenced by them.
Uncertain respondents are closer to Iraquí respondents, Shik, and those who did not respond to the gender question.
2.5 Clustering
covidVax.MCA$eig #%>% head() eigenvalue percentage of variance cumulative percentage of variance
dim 1 0.4200437 13.089489 13.08949
dim 2 0.2632400 8.203140 21.29263
dim 3 0.2327096 7.251746 28.54437
dim 4 0.2160854 6.733700 35.27807
dim 5 0.2076532 6.470933 41.74901
dim 6 0.2075218 6.466837 48.21585
dim 7 0.2004154 6.245388 54.46123
dim 8 0.1987885 6.194688 60.65592
dim 9 0.1911314 5.956075 66.61200
dim 10 0.1797189 5.600437 72.21243
dim 11 0.1777702 5.539713 77.75215
dim 12 0.1700613 5.299485 83.05163
dim 13 0.1594057 4.967434 88.01907
dim 14 0.1467724 4.573751 92.59282
dim 15 0.1222062 3.808214 96.40103
dim 16 0.1154914 3.598968 100.00000Choosing 14 dimension to get >90% of variance
Performing clustering
# Run this in console to choose the number of clusters
# covidVax.hcpc <- HCPC(covidVax.MCA)
# Leave this on chunk for rendering
covidVax.hcpc <- HCPC(covidVax.MCA, nb.clust = 3) 
This plot identifies 3 clusters, suggested by dendogram
# plot.HCPC(res.HCPC,choice='map',draw.tree=FALSE,title='Factor map', label = "none")
plot(covidVax.hcpc,choice="map",ind.names=FALSE)
round(covidVax.hcpc$call$t$inert.gain,3) %>% head()[1] 0.235 0.145 0.121 0.100 0.063 0.062round(covidVax.MCA$eig[,1],3) %>% head()dim 1 dim 2 dim 3 dim 4 dim 5 dim 6
0.420 0.263 0.233 0.216 0.208 0.208 Defining the clusters
covidVax.hcpc$desc.var$test p.value df
leadOpCVax 0.000000e+00 6
mostLeadEndorseCV 0.000000e+00 4
vaxed 0.000000e+00 2
influencerCVax 0.000000e+00 8
factorCVax 0.000000e+00 14
country 0.000000e+00 18
role 1.901006e-211 4
religion 3.317285e-188 18
contact_type 2.362398e-138 2
sex 1.157525e-128 4
age 6.615220e-19 6
geoLoc 1.247335e-03 2Many variables appear relevant to define the cluster, mostly the importance of leader’s opinion on Covid vaccination, the view on leader’s endorsement of covid vaccination, the vaccination status, the choice of an influencer on covid vaccination, the influencing factor, and the country.
Categories show
Cluster 1:
- Those who believe the most important factor for decisions on COVID vaccination is seeing their religious leader taking the vaccine
- Those who feel most influenced by health workers
- Those who are vaccinated
- Those who believe their religious leader endorse COVID vaccination
- Those who believe their leader's opinion matters to themCluster 2:
- Those who believe the most important factor for decisions on COVID vaccination is the media
- Those who feel most influenced by peers and familyCluster 3:
- Those who feel most influenced by other people, and by other factors
- Those who are not vaccinated
- Those unsure of their leader's opinion on COVID vaccination and those who are not sure if that opinion matters to themcovidVax.hcpc$desc.var$category %>% lapply(head)$`1`
Cla/Mod Mod/Cla Global p.value
factorCVax=fLeaderTakingVax 89.21833 33.96480 24.30213 0.000000e+00
influencerCVax=infCvax_hWorker 84.02150 50.59594 38.44100 0.000000e+00
vaxed=vaxed 72.86085 86.50249 75.78857 0.000000e+00
mostLeadEndorseCV=mostLeadEndorseCVax 80.70314 80.81143 63.92220 0.000000e+00
leadOpCVax=leadOpCVax_matters 85.30351 56.90268 42.58289 0.000000e+00
factorCVax=fClearInfo 78.72920 43.32623 35.13050 5.247104e-237
v.test
factorCVax=fLeaderTakingVax Inf
influencerCVax=infCvax_hWorker Inf
vaxed=vaxed Inf
mostLeadEndorseCV=mostLeadEndorseCVax Inf
leadOpCVax=leadOpCVax_matters Inf
factorCVax=fClearInfo 32.87356
$`2`
Cla/Mod Mod/Cla Global
factorCVax=fMedia 100.00000 48.81398 5.910511
influencerCVax=infCvax_peerFriend 93.99460 57.96921 7.467500
leadOpCVax=leadOpCVax_ns 16.84947 30.04578 21.591253
mostLeadEndorseCV=notSureLeadEndorseCVax 16.38966 29.54640 21.828076
role=notPart 14.70221 44.06991 36.294467
contact_type=RDD / 3-2-1 13.28892 78.85976 71.853270
p.value v.test
factorCVax=fMedia 0.000000e+00 Inf
influencerCVax=infCvax_peerFriend 0.000000e+00 Inf
leadOpCVax=leadOpCVax_ns 3.080650e-25 10.379288
mostLeadEndorseCV=notSureLeadEndorseCVax 2.863931e-21 9.467583
role=notPart 6.186671e-17 8.361613
contact_type=RDD / 3-2-1 6.929390e-17 8.348229
$`3`
Cla/Mod Mod/Cla Global
factorCVax=fOther 93.65918 29.70256 7.628741
influencerCVax=infCvax_other 68.04071 56.01173 19.802479
vaxed=unvaxed 48.99063 49.30876 24.211428
mostLeadEndorseCV=notSureLeadEndorseCVax 66.25115 60.11730 21.828076
leadOpCVax=leadOpCVax_ns 61.02684 54.77587 21.591253
role=notPart 34.79106 52.49267 36.294467
p.value v.test
factorCVax=fOther 0.000000e+00 Inf
influencerCVax=infCvax_other 0.000000e+00 Inf
vaxed=unvaxed 0.000000e+00 Inf
mostLeadEndorseCV=notSureLeadEndorseCVax 0.000000e+00 Inf
leadOpCVax=leadOpCVax_ns 0.000000e+00 Inf
role=notPart 5.920643e-153 26.34459