forked from elmoen/clustered_data_code
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Stata
92 lines (63 loc) · 2.95 KB
/
Stata
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
Stata code for statistical models described in the main text
Key
Variable names:
somasize = continuous variable of soma size for each neuron
fattyacid = factor of 0, 1, or 2 indicating no treatment, vehicle control, or fatty acid treatment respectively
pten = factor of 0 or 1 indicating control or Pten shRNA, respectively
mouseid = factor specifying mouse identity for each neuron
meanss = mean soma size
prp = proportion of cells with Pten knockdown per mouse
numobs = number of neuron measurements per mouse
Code corresponding to Table 3
Experimental subset: control neurons from mice that received fatty acid or vehicle control delivery (pten = 0 and fattyacid = 1 or 2)
1) Cell-level linear regression
reg somasize fattyacid
2) Mouse-level regression (mean); no weighting
reg meanss fattyacid
3) Mouse-level regression (mean); weighting – analytic weights
reg meanss fattyacid [w=numobs]
4) Marginal regression
#First set mouseid as the panel variable (cluster identities), then perform the regression
xtset mouseid
xtgee somasize fattyacid, corr(exch) robust
5) Mixed-effect regression
#Use of “variance robust” obtains variance estimates that are robust to violations of the distribution assumption due to misspecification of the model
xtset mouseid
xtmixed somasize fattyacid || mouseid:, variance robust cov(exch)
Stata code corresponding to Table 4
Experimental subset: mice that remained naïve to either fatty acid or vehicle control delivery (fattyacid = 0)
1) Cell-level linear regression
reg somasize pten
2) Marginal regression
xtset mouseid
xtgee somasize pten, corr(exch) robust
3) Fixed-effect regression
xtset mouseid
xtreg somasize pten, fe
4) Mixed-effect regression
xtset mouseid
xtmixed somasize pten || mouseid:, variance robust cov(exch)
Stata code corresponding to Table 5
Experimental subset: mice that remained naïve to either fatty acid or vehicle control delivery (fattyacid = 0)
1) Mouse-level weighted regression
reg meanss prp [w=numobs]
2) Marginal regression
xtset mouseid
xtgee somasize prp pten, corr(exch) robust
3) Mixed-effect regression
xtset mouseid
xtmixed somasize prp ||mouseid:, variance robust cov(exch)
4) Mixed-effect regression with pten as a fixed effect
xtset mouseid
xtmixed somasize prp pten || mouseid:, variance robust cov(exch)
Stata code corresponding to Table 6
Experimental subset: mice that were in the fatty acid and vehicle control groups (fattyacid = 1 or 2).
1) Interaction effect between Pten knockdown and fatty acid delivery
#Set mouseid as panel variable, generate vectors of interaction effects, and then perform the regression
xtset mouseid
gen fa1_pten=(fattyacid==1)*pten
gen fa2_pten=(fattyacid==2)*pten
xtmixed somasize pten fattyacid fa1_pten fa2_pten || mouseid:, variance robust cov(exch)
Intra-class correlation (ICC)
#After performing the mixed-effect model with the ‘xtmixed’ command, you can calculate the ICC of your data with the following command:
estat icc