-
Notifications
You must be signed in to change notification settings - Fork 0
/
sexual_violence_analysisv3.m
270 lines (253 loc) · 11.9 KB
/
sexual_violence_analysisv3.m
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
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
load filtered_db_2015_02_24
load GENDER
%also load some deomgraphic analysis !
%2. SEXUAL VIOLENCE/EXCHANGE VARIABLES
%%--these first variables are in the "exchange" and "exchange_N" variables
% % for total N for all variables; frequencies, percentages and p-values
% % for all variables by gender.
% % Touched/rubbed in sexual way (sexvio_Note1/sexvio_1) ? this looks correct (binary)
% % Touched/rubbed in sexual way # of times (sexvio_Note1/sexvio_2) ? this looks correct (continuous)
% % Collapse into 3 groups: 1, 2-4, 5 or more
% %
% % Fingers/objects inserted (sexvio_Note1/sexvio_3) ? this looks correct (binary)
% % Fingers/objects inserted # of times (sexvio_Note1/sexvio_4) ? this looks correct (continuous)
% % Collapse into 3 groups: 1, 2-4, 5 or more
% %
% % Sex without consent (sexvio_Note1/sexvio_5) ? this looks correct (binary)
% % Sex without consent # of times (sexvio_Note1/sexvio_6) ? this looks correct (continuous)
% % Collapse into 3 groups: 1, 2-4, 5 or more
% %
% % Felt sex expected because using together (sexvio_Note1/sexvio_9) ? this looks correct (binary)
% % Felt sex expected because using together # of times (sexvio_Note1/sexvio_10) ? this looks correct (continuous), Collapse into 3 groups: 1-4, 5-9, 10 or more
% % Sexually insulted (sexvio_Note1/sexvio_11) ? this looks correct (binary)
% % Sexually insulted # of times (sexvio_Note1/sexvio_12) ? this looks correct (continuous),
% % Collapse into 3 groups: 1-4, 5-9, 10 or more
% %
% % Offered drugs/$ for sex (sexvio_Note1/sexvio_13) ? this looks correct (binary)
% % Offered drugs/$ for sex # of times (sexvio_Note1/sexvio_14) ? this looks correct (continuous),
% % Frequency & % for following groups: 1-4, 5-9, 10 or more, 50 or more
% % Felt sexually violated but don't remember (sexvio_Note1/sexvio_15) ? this looks correct (binary)
% %
% % Witnessed sexual violence? this looks correct (binary, as TRUE/FALSE right now)
% % sexvio_Note1/sexvio_16/1 -- witnessed someone touching/rubbing in a
% sexual way
% % sexvio_Note1/sexvio_16/2 --witnessed oral sex without consent
% % sexvio_Note1/sexvio_16/3 --witnessed finger/object insertion
% % sexvio_Note1/sexvio_16/4 -- witnessed sex without consent
% % sexvio_Note1/sexvio_16/88 --Don't know <----DON'T DO THIS
% %
% % Sexual exchange (LifetimeSex_3) ? this looks correct (binary)
% % Sexual exchange received drugs/$ # times (LifetimeSex_4) ? this looks correct (continuous)
% % Collapse into 3 groups: 1, 2-4, 5 or more
% %
% % Sexual exchange paid drugs/$ # times (LifetimeSex_5) ? this looks correct (continuous)
% % Collapse into 3 groups: 1, 2-4, 5 or more
% %
sex_exchanged={'LifetimeSex_3', 'LifetimeSex_4', 'LifetimeSex_5'};
%THE LIFETIME 5 doesn't have an intital binary
names={'Touched/rubbed in sexual way', 'Fingers/objects inserted', 'Sex without consent', 'Felt sex expected because using together', 'Sexually insulted', 'Offered drugs/$ for sex', 'Felt sexually violated but does not remember', 'Witnessed someone touching/rubbing in sexual way', 'Witnessed oral sex without consent', 'Witnessed finger/object insertion', 'Witnessed sex without consent' 'Sexual exchange received drugs/$', 'Sexual exchange paid drugs/$' };
exchange={'sexvio_Note1/sexvio_1','sexvio_Note1/sexvio_3', 'sexvio_Note1/sexvio_5', 'sexvio_Note1/sexvio_9','sexvio_Note1/sexvio_11', 'sexvio_Note1/sexvio_13', 'sexvio_Note1/sexvio_15', 'sexvio_Note1/sexvio_16/1', 'sexvio_Note1/sexvio_16/2', 'sexvio_Note1/sexvio_16/3', 'sexvio_Note1/sexvio_16/4', 'LifetimeSex_3', 'LifetimeSex_3'};
exchange_N={'sexvio_Note1/sexvio_2','sexvio_Note1/sexvio_4','sexvio_Note1/sexvio_6', 'sexvio_Note1/sexvio_10','sexvio_Note1/sexvio_12','sexvio_Note1/sexvio_14', '', '', '', '', '', 'LifetimeSex_4', 'LifetimeSex_5' };
[r,c]=size(filtered_final);
filtered_data=filtered_final(2:r,:);
headers=filtered_final(1,:);
results=cell.empty;
results{1,1}='Question';
results{1,2}='Code';
results{1,3}='freq-total';
results{1,4}='% total';
results{1,5}='95% CI';
results{1,6}='freq-males';
results{1,7}='% males';
results{1,8}='95% CI';
results{1,9}='freq-females';
results{1,10}='% females';
results{1,11}='95% CI';
results{1,12}='p value';
T={'1', '2-4', '5-9', '10-50' '>50', '1 %', '2-4 %', '5-9 %', '10-50 %', '>50 %', '1', '2-4', '5-9', '10-50' '>50', '1 %', '2-4 %', '5-9 %', '10-50 %', '>50 %', '1', '2-4', '5-9', '10-50' '>50', '1 %', '2-4 %', '5-9 %', '10-50 %', '>50 %'};
results(1, 12:41)=T;
z=1.96;
for i=1:numel(exchange)
E=exchange{i};
indx=find(strcmp(headers,E)==1);
results{i+1,1}=names{i};
E_N=exchange_N{i};
indx_N=find(strcmp(headers,E_N)==1);
n_check=0;
if numel(indx)==1
data_mat=filtered_data(:,indx);
indx_nan=find(strcmp('NaN', data_mat)==1);
for j=1:numel(indx_nan)
data_mat{indx_nan(j)}=NaN;
end
tf=isa(data_mat,'cell');
indx_t=find(strcmp('TRUE',data_mat)==1);
if numel(indx_t)>0%tf==1
data_mat2=NaN(size(data_mat));
indx_t=find(strcmp('TRUE',data_mat)==1);
indx_f=find(strcmp('FALSE',data_mat)==1);
for j=1:numel(indx_t)
data_mat2(indx_t(j))=1;
end
for j=1:numel(indx_f)
data_mat2(indx_f(j))=0;
end
% indx_empty=cellfun('isempty',data_mat);
% data_mat{indx_empty}=NaN;
% indx_nan=find(strcmp('NaN',data_mat)==1);
% for j=1:numel(indx_nan)
% data_mat{indx_nan(j)}=NaN;
% end
% isnum = cellfun(@isnumeric,data_mat);
% data_mat2 = NaN(size(data_mat));
% data_mat2(isnum) = [data_mat{isnum}];
data_mat=data_mat2;
else
data_mat=cell2mat(data_mat);
end
indx_male=find(GENDER==1);
indx_female=find(GENDER==2);
data_=find(isnan(data_mat)==0);
if numel(indx_N)==1
results{i+1,2}=[E '-' E_N];
data_n=filtered_data(:,indx_N);
indx_nan=find(strcmp('NaN', data_n)==1);
for j=1:numel(indx_nan)
data_n{indx_nan(j)}=NaN;
end
data_n=cell2mat(data_n);
n_check=1;
else
results{i+1,2}=E;
end
%TOTAL------------------------------
N=numel(data_); %total answered
N_m=numel(intersect(data_,indx_male));
N_f=numel(intersect(data_,indx_female));
if (n_check==1)
indx_yes=find(data_mat==1 & data_n>0);
else
indx_yes=find(data_mat==1);
end
results{i+1,3}=numel(indx_yes);
P=numel(indx_yes)/N;
upper=((P+z*sqrt(P*(1-P)/N))*100);
lower=((P-z*sqrt(P*(1-P)/N))*100) ;
p=P*100;
results{i+1,4}=[sprintf('%0.1f',p) '%'];
upper=sprintf('%0.1f',round(upper*10)/10);
lower=sprintf('%0.1f',round(lower*10)/10);
results{i+1,5}=[lower '%-' upper '%'];
%Deal with the N
if n_check==1
indx_N=find(data_n>0);
indx_1=find (data_n==1);
indx_2=find(data_n>1 & data_n<5);
indx_5=find(data_n>4 & data_n<10);
indx_10=find(data_n>9 & data_n<50);
indx_50=find(data_n>49);
results{i+1,12}=numel(indx_1);
results{i+1,13}=numel(indx_2);
results{i+1,14}=numel(indx_5);
results{i+1,15}=numel(indx_10);
results{i+1,16}=numel(indx_50);
p_1=numel(indx_1)/numel(indx_N)*100;
p_2=numel(indx_2)/numel(indx_N)*100;
p_5=numel(indx_5)/numel(indx_N)*100;
p_10=numel(indx_10)/numel(indx_N)*100;
p_50=numel(indx_50)/numel(indx_N)*100;
results{i+1,17}=[sprintf('%.1f',p_1) '%'];
results{i+1,18}=[sprintf('%.1f',p_2) '%'];
results{i+1,19}=[sprintf('%.1f',p_5) '%'];
results{i+1,20}=[sprintf('%.1f',p_10) '%'];
results{i+1,21}=[sprintf('%.1f',p_50) '%'];
end
%MALE-------------------------------
indx_M=intersect(indx_male,indx_yes);
results{i+1,6}=numel(indx_M);
P=numel(indx_M)/N_m;
upper=((P+z*sqrt(P*(1-P)/N_m))*100);
lower=((P-z*sqrt(P*(1-P)/N_m))*100) ;
p=P*100;
results{i+1,7}=[sprintf('%0.1f',p) '%'];
upper=sprintf('%0.1f',round(upper*10)/10);
lower=sprintf('%0.1f',round(lower*10)/10);
results{i+1,8}=[lower '%-' upper '%'];
if n_check==1
indx_M_N=intersect(indx_male,indx_N);
indx_1=find (data_n==1);
indx_2=find(data_n>1 & data_n<5);
indx_5=find(data_n>4 & data_n<10);
indx_10=find(data_n>9 & data_n<50);
indx_50=find(data_n>49);
indx_1_=intersect(indx_1,indx_male);
indx_2_=intersect(indx_2,indx_male);
indx_5_=intersect(indx_5,indx_male);
indx_10_=intersect(indx_10,indx_male);
indx_50_=intersect(indx_50, indx_male);
results{i+1,22}=numel(indx_1_);
results{i+1,23}=numel(indx_2_);
results{i+1,24}=numel(indx_5_);
results{i+1,25}=numel(indx_10_);
results{i+1,26}=numel(indx_50_);
p_1=numel(indx_1_)/numel(indx_M_N)*100;
p_2=numel(indx_2_)/numel(indx_M_N)*100;
p_5=numel(indx_5_)/numel(indx_M_N)*100;
p_10=numel(indx_10_)/numel(indx_M_N)*100;
p_50=numel(indx_50_)/numel(indx_M_N)*100;
results{i+1,27}=[sprintf('%.1f',p_1) '%'];
results{i+1,28}=[sprintf('%.1f',p_2) '%'];
results{i+1,29}=[sprintf('%.1f',p_5) '%'];
results{i+1,30}=[sprintf('%.1f',p_10) '%'];
results{i+1,31}=[sprintf('%.1f',p_50) '%'];
end
indx_F=intersect(indx_female,indx_yes);
results{i+1,9}=numel(indx_F);
P=numel(indx_F)/N_f;
upper=((P+z*sqrt(P*(1-P)/N_f))*100);
lower=((P-z*sqrt(P*(1-P)/N_f))*100) ;
p=P*100;
results{i+1,10}=[sprintf('%.1f',p) '%'];
upper=sprintf('%0.1f',round(upper*10)/10);
lower=sprintf('%0.1f',round(lower*10)/10);
results{i+1,11}=[lower '%-' upper '%'];
if n_check==1
indx_F_N=intersect(indx_female,indx_N);
indx_1=find (data_n==1);
indx_2=find(data_n>1 & data_n<5);
indx_5=find(data_n>4 & data_n<10);
indx_10=find(data_n>9 & data_n<50);
indx_50=find(data_n>49);
indx_1_=intersect(indx_1,indx_female);
indx_2_=intersect(indx_2,indx_female);
indx_5_=intersect(indx_5,indx_female);
indx_10_=intersect(indx_10,indx_female);
indx_50_=intersect(indx_50, indx_female);
results{i+1,32}=numel(indx_1_);
results{i+1,33}=numel(indx_2_);
results{i+1,34}=numel(indx_5_);
results{i+1,35}=numel(indx_10_);
results{i+1,36}=numel(indx_50_);
p_1=numel(indx_1_)/numel(indx_F_N)*100;
p_2=numel(indx_2_)/numel(indx_F_N)*100;
p_5=numel(indx_5_)/numel(indx_F_N)*100;
p_10=numel(indx_10_)/numel(indx_F_N)*100;
p_50=numel(indx_50_)/numel(indx_F_N)*100;
results{i+1,37}=[sprintf('%.1f',p_1) '%'];
results{i+1,38}=[sprintf('%.1f',p_2) '%'];
results{i+1,39}=[sprintf('%.1f',p_5) '%'];
results{i+1,40}=[sprintf('%.1f',p_10) '%'];
results{i+1,41}=[sprintf('%.1f',p_50) '%'];
end
x=max(numel(indx_male),numel(indx_female));
stat_mat=nan(x,2);
stat_mat(1:numel(indx_male),1)=zeros(numel(indx_male),1);
stat_mat(1:numel(indx_female),2)=zeros(numel(indx_female),1);
stat_mat(1:numel(indx_M),1)=ones(1,numel(indx_M));
stat_mat(1:numel(indx_F),2)=ones(1,numel(indx_F));
[h,p]=ttest2(stat_mat(:,1), stat_mat(:,2));
results{i+1,42}=p;
else
E
end
end