SPINS participant matching:
preparation for DWI harmonization
data pulled: 2019-05-20
written: 2019-05-25
last ran: 2019-05-26
Overview
Purpose. Our aim is to match participants across the 6 SPINS sites, in keeping with Karayumak et al., 2019. More specifically, in preparation of harmonizing the raw DWI data, we first aim to identify the gold-standard of n=20 matched ‘healthy control’ participants (or at least, the recommended n=16-18), on the basis of handedness (right; here, established by EHI ≥ .5), sex, age, and IQ (here, established by WTAR). We have decided that CMH (i.e., the GE Discovery scanner in Toronto) will be the ‘reference’ site, to which the other 5 ‘target’ sites will be harmonized.
Data cleaning. We removed healthy control participants who (i) did not pass diffusion / T1 QC, (ii) were not right-handed, and (iii) did not have complete sex, age, and/or WTAR data. After this winnowing, we see that we have just enough participants to attempt matching of CMH to MRC, MRP, and ZHP.
reference
target
CMH
CMP
MRC
MRP
ZHH
ZHP
female
16
4
9
10
5
17
male
23
3
13
11
7
17
Matching method. We performed matching with the MatchIt R package, using a one-to-one nearest neighbours approach (i.e., each participant in the target dataset was optimally matched with a participant in the reference dataset). Due to non-replacement, the number of matches is synonymous with the smallest n per sex across sites (whether belonging to the reference or target site). Then, we selected the 20 best matches (10 per sex selected if possible), as determined by smallest absolute value of matching distance (see Guo & Fraser, 2015, for details of this distance calculation). In the ‘Matching results’ section below, we display distance values, as well as match-adjusted mean values for age and WTAR variables.
Dataset descriptions
MRC
There is a significant difference in age between sites, in both Males and Females. There are no differences in IQ or sex.
CMH
MRC
p
test
n
39
22
age (mean (SD))
27.08 (8.34)
36.95 (11.05)
<0.001
sex = male (%)
23 (59.0)
13 (59.1)
1.000
wtar_std (mean (SD))
115.33 (9.39)
115.86 (10.22)
0.838
Males
CMH
MRC
p
test
n
23
13
age (mean (SD))
26.78 (8.23)
35.08 (9.19)
0.009
wtar_std (mean (SD))
115.35 (9.20)
117.85 (7.26)
0.406
Females
CMH
MRC
p
test
n
16
9
age (mean (SD))
27.50 (8.75)
39.67 (13.41)
0.011
wtar_std (mean (SD))
115.31 (9.95)
113.00 (13.40)
0.627
MRP
There is a significant difference in age, driven by Females, and IQ, driven by Males, between sites. There is no difference in sex.
CMH
MRP
p
test
n
39
21
age (mean (SD))
27.08 (8.34)
34.29 (11.79)
0.008
sex = male (%)
23 (59.0)
11 (52.4)
0.827
wtar_std (mean (SD))
115.33 (9.39)
109.19 (11.28)
0.028
Males
CMH
MRP
p
test
n
23
11
age (mean (SD))
26.78 (8.23)
31.73 (11.95)
0.167
wtar_std (mean (SD))
115.35 (9.20)
107.27 (13.20)
0.046
Females
CMH
MRP
p
test
n
16
10
age (mean (SD))
27.50 (8.75)
37.10 (11.54)
0.024
wtar_std (mean (SD))
115.31 (9.95)
111.30 (8.92)
0.309
ZHP
There is a significant difference in age between sites, and this is driven by Males. There are no differences in IQ or sex.
CMH
ZHP
p
test
n
39
34
age (mean (SD))
27.08 (8.34)
35.26 (11.18)
0.001
sex = male (%)
23 (59.0)
17 (50.0)
0.594
wtar_std (mean (SD))
115.33 (9.39)
112.79 (12.10)
0.317
Males
CMH
ZHP
p
test
n
23
17
age (mean (SD))
26.78 (8.23)
39.71 (11.08)
<0.001
wtar_std (mean (SD))
115.35 (9.20)
115.35 (11.59)
0.999
Females
CMH
ZHP
p
test
n
16
17
age (mean (SD))
27.50 (8.75)
30.82 (9.65)
0.309
wtar_std (mean (SD))
115.31 (9.95)
110.24 (12.39)
0.206
Matching results
MRC
13 Male matches were made (min value = MRC), and 9 Female matches were made (min value = MRC). We see that, after matching on n=20 (11 Males, 9 Females), age is no longer significantly different (unpaired t-test) between CMH and MRC.
CMH
MRC
p
test
n
20
20
age (mean (SD))
30.75 (9.99)
36.15 (11.17)
0.115
sex = male (%)
11 (55.0)
11 (55.0)
1.000
wtar_std (mean (SD))
114.65 (11.16)
115.80 (10.38)
0.738
distance (mean (SD))
0.37 (0.22)
0.50 (0.23)
0.069
MRP
11 Male matches were made (min value = MRP), and 10 Female matches were made (min value = MRP). We see that, after matching on n=20 (10 Males, 10 Females), neither age nor WTAR remains significantly different (unpaired t-test) between CMH and MRP.
CMH
MRP
p
test
n
20
20
age (mean (SD))
29.85 (9.24)
34.75 (11.89)
0.154
sex = male (%)
10 (50.0)
10 (50.0)
1.000
wtar_std (mean (SD))
112.35 (9.49)
110.30 (10.33)
0.517
distance (mean (SD))
0.37 (0.18)
0.46 (0.22)
0.179
ZHP
17 Male matches were made (min value = ZHP), and 16 Female matches were made (min value = CMH). We see that, after matching on n=20 (10 Males, 10 Females), age is no longer significantly different (unpaired t-test) between CMH and ZHP.