-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset5_contradiction.R
222 lines (152 loc) · 7.54 KB
/
dataset5_contradiction.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
library(mirt)
library(psych)
require(graphics)
answers_5 <- read.csv("irt_file_5_coded.csv",header=TRUE,sep=",")
Key<-unclass(read.table("group_5.txt",sep=" ",comment.char="")[,2]);
snli_5 <- answers_5[2:91]
c_s5 <- colnames(snli_5)
Data <- snli_5;
items<-which(Key==1); #key 1 is contradiction
#need to remove the items that were marked via visual inspection
which(c_s5[items] == 'X5794490455.jpg.0r1c')
which(c_s5[items] == 'X4814925710.jpg.0r1c')
#remove items 22 and 26 for first run
c_s5[items]
######################
### Dimensionality ###
######################
R<-tetrachoric(Data[,items[c(-22,-26)]])
plot(eigen(R$rho,symmetric=TRUE,only.values=TRUE)$values)
### One single high eigenvalue before trailing eigenvalues, suggesting a single factor
fitMIRT1<-mirt(Data[,items[c(-22,-26)]],1,itemtype="3PL",TOL=1e-4,technical=list(NCYCLES=1e4))
fitMIRT2<-mirt(Data[,items[c(-22,-26)]],2,itemtype="3PL",TOL=1e-4,technical=list(NCYCLES=1e4))
anova(fitMIRT1,fitMIRT2) ## Sample size adjusted BIC suggests 2F model
summary(fitMIRT2,suppress=0.3) ## 2F model shows split between two factors
# new rotation to sqeeze into single factor
target<-cbind(rep(1,28),rep(0,28));
summary(fitMIRT2,suppress=0.3,rotate="pstQ",Target=target,W=!target) #now we see small loading for 2nd factor
#This suggests single factor.
#######################
## Evaluate 1F model ##
#######################
fit.all.est<-coef(fitMIRT1,simplify=TRUE,IRTpar=TRUE)$items;
mtlnL.all<- -2*fitMIRT1@Fit$logLik;
pvalue3v2<-rep(NA,28);
type<-rep("3PL",28);
for (k in 1:28)
{
if (fit.all.est[k,"g"]<1e-2) {pvalue3v2[k]<-.5; next;}
itemtype<-type;
itemtype[k]<-"2PL";
fit.2PL <-mirt(Data[,items[c(-22,-26)]],1,itemtype,TOL=2e-4,technical=list(NCYCLES=1e4))
mtlnL.2PL<- -2*fit.2PL@Fit$logLik;
T3v2<-mtlnL.2PL-mtlnL.all;
pvalue3v2[k]<-p<-pchisq(T3v2,df=1,lower.tail=FALSE)/2;
cat(items[k],": 3PL vs 2PL ",T3v2," p = ", p, "\n");
}
s<-sort.int(pvalue3v2,index.return=TRUE)
print(cbind(items[s$ix],s$x,0.05/(28:1))); #holm criteria
## No lower asymptote is significant. Should change to 2PL.
fit.2PL<-mirt(Data[,items[c(-22)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
anova(fit.2PL,fitMIRT1) ## Test confirm 2PL model is as good as 3PL.
summary(fit.2PL) ## There are three items with h2 (communality) smaller than 15%. They should be removed.
coef(fit.2PL,simplify=TRUE,IRTpar=TRUE)$items;
# 9, 10, 14 should be removed
#############################
### Remove Items 9, 10, and 14 ###
#############################
fit.2<-mirt(Data[,items[c(-9,-10,-14,-22,-26)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are significant, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(25:1)) #one value is significant, item 9 (original item 11)
res.table<-residuals(fit.2)
pstar<-25*24/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# several significant pairs here
#now remove item 11
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-22,-26)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are significant, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
summary(fit.2)
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(24:1)) # one is significant, item 24 (originally item 30)
res.table<-residuals(fit.2)
pstar<-24*23/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# 3 significant pairs found.
# next remove item 30
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-22,-26,-30)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are significant, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(23:1)) # one is significant, item 22 (originally item 28)
res.table<-residuals(fit.2)
pstar<-23*22/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# 2 significant pairs found.
# remove item 28
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-22,-26,-28,-30)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are significant, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(22:1)) # none are significant
res.table<-residuals(fit.2)
pstar<-22*21/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# still have 2 significant pairs found.
residuals(fit.2,df.p=TRUE,suppress=0.13)
residuals(fit.2)
# highest LD value between X2326730558.jpg.4r2c and X630476551.jpg.4r1c
# items 20 & 19
# which has lower communality?
summary(fit.2) #they're about the same, so we'll try each model.
# remove item 19
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-19,-22,-26,-28,-30)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are below 0.05, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(21:1)) # none are significant
res.table<-residuals(fit.2)
pstar<-21*20/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# still have 1 significant pair found.
#now let's try without 20
# remove item 28
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-20,-22,-26,-28,-30)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## some p values are less than 0.05, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(21:1)) # none are significant
res.table<-residuals(fit.2)
pstar<-21*20/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# still have 1 significant pairs found.
#since removing each left one significant value, remove both
fit.2<-mirt(Data[,items[c(-9,-10,-11,-14,-19,-20,-22,-26,-28,-30)]],1,itemtype="2PL",TOL=2e-4,technical=list(NCYCLES=1e3))
itf <- itemfit(fit.2) ## one p value less than 0.05, we'll need to test them with Holm
M2(fit.2) ## RMSEA_5 smaller than 0.5; TLI and CFI greater than 0.95. Good overall fit for 2-margins.
#first test p values of itemfit using Holm criteria
cbind(sort(itf$p.S_X2),0.05/(20:1)) # none are significant
# uncomment to check original item order
# c_s5[items]
res.table<-residuals(fit.2)
pstar<-20*19/2
t <- 1-pchisq(df=1,abs(res.table[lower.tri(res.table)]))
cbind(sort(round(t,6)),0.05/(pstar:1))
## Compare the p-values to the Holm criteria.
# no significant pairs found.
# set of good items from this analysis:
# 1,2,3,4,5,6,7,8,12,13,15,16,17,18,21,23,24,25,27,29