Package Comparison Details, by dataset
Cranio Dataset
Soft Threshold Choice
Comparing the Sage code base to the UCLA-based package, we see a difference in the selection of the 'soft threshold' (beta):
Sage power table / UCLA power table (The optimum chosen by each is highlighted)
|
Power |
p-value |
Adj R^2 |
Truncated Adj R^2 |
slope |
mean(k) |
median(k) |
max(k) |
|
|
Power |
SFT.R.sq |
slope |
truncated.R.sq |
mean.k. |
median.k. |
max.k. |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 |
1 |
-1 |
0.707 |
0.991 |
1.18 |
700 |
700 |
1120 |
|
1 |
1 |
0.737 |
1.14 |
0.994 |
700 |
700 |
1120 |
2 |
1.5 |
-1 |
0.231 |
0.969 |
0.332 |
440 |
424 |
840 |
|
2 |
1.5 |
0.252 |
0.298 |
0.979 |
440 |
424 |
840 |
3 |
2 |
-1 |
0.0114 |
0.903 |
-0.136 |
296 |
269 |
668 |
|
3 |
2 |
0.12 |
-0.164 |
0.917 |
296 |
269 |
668 |
4 |
2.5 |
-1 |
0.523 |
0.921 |
-0.444 |
209 |
178 |
549 |
|
4 |
2.5 |
0.623 |
-0.491 |
0.901 |
209 |
178 |
549 |
5 |
3 |
-1 |
0.749 |
0.914 |
-0.66 |
153 |
120 |
463 |
|
5 |
3 |
0.788 |
-0.683 |
0.904 |
153 |
120 |
463 |
6 |
3.5 |
-1 |
0.819 |
0.892 |
-0.798 |
115 |
83.7 |
399 |
|
6 |
3.5 |
0.863 |
-0.857 |
0.889 |
115 |
83.7 |
399 |
7 |
4 |
-1 |
0.864 |
0.892 |
-0.918 |
89.5 |
59.4 |
351 |
|
7 |
4 |
0.895 |
-0.952 |
0.894 |
89.5 |
59.4 |
351 |
8 |
4.5 |
-1 |
0.884 |
0.888 |
-1.01 |
70.9 |
42.2 |
313 |
|
8 |
4.5 |
0.906 |
-1.04 |
0.887 |
70.9 |
42.2 |
313 |
9 |
5 |
-1 |
0.879 |
0.872 |
-1.08 |
57.3 |
30.5 |
283 |
|
9 |
5 |
0.917 |
-1.09 |
0.894 |
57.3 |
30.5 |
283 |
10 |
5.5 |
-1 |
0.887 |
0.878 |
-1.12 |
47.1 |
22.6 |
258 |
|
10 |
5.5 |
0.903 |
-1.13 |
0.876 |
47.1 |
22.6 |
258 |
11 |
6 |
-1 |
0.872 |
0.864 |
-1.14 |
39.3 |
16.8 |
237 |
|
11 |
6 |
0.894 |
-1.14 |
0.872 |
39.3 |
16.8 |
237 |
12 |
6.5 |
-1 |
0.853 |
0.849 |
-1.16 |
33.2 |
12.5 |
220 |
|
12 |
6.5 |
0.877 |
-1.17 |
0.861 |
33.2 |
12.5 |
220 |
13 |
7 |
-1 |
0.849 |
0.857 |
-1.16 |
28.5 |
9.75 |
205 |
|
13 |
7 |
0.868 |
-1.17 |
0.863 |
28.5 |
9.75 |
205 |
14 |
7.5 |
-1 |
0.836 |
0.861 |
-1.16 |
24.7 |
7.44 |
193 |
|
14 |
7.5 |
0.869 |
-1.16 |
0.888 |
24.7 |
7.44 |
193 |
15 |
8 |
-1 |
0.831 |
0.873 |
-1.15 |
21.6 |
5.73 |
182 |
|
15 |
8 |
0.86 |
-1.15 |
0.89 |
21.6 |
5.73 |
182 |
16 |
8.5 |
-1 |
0.807 |
0.87 |
-1.15 |
19.1 |
4.43 |
172 |
|
16 |
8.5 |
0.84 |
-1.15 |
0.89 |
19.1 |
4.43 |
172 |
17 |
9 |
-1 |
0.78 |
0.858 |
-1.14 |
17 |
3.49 |
164 |
|
17 |
9 |
0.828 |
-1.13 |
0.896 |
17 |
3.49 |
164 |
18 |
9.5 |
-1 |
0.792 |
0.889 |
-1.11 |
15.3 |
2.75 |
156 |
|
18 |
9.5 |
0.825 |
-1.1 |
0.912 |
15.3 |
2.75 |
156 |
19 |
10 |
-1 |
0.782 |
0.906 |
-1.09 |
13.9 |
2.18 |
149 |
|
19 |
10 |
0.806 |
-1.09 |
0.915 |
13.9 |
2.18 |
149 |
20 |
10.5 |
-1 |
0.759 |
0.897 |
-1.07 |
12.7 |
1.74 |
143 |
|
20 |
10.5 |
0.8 |
-1.08 |
0.921 |
12.7 |
1.74 |
143 |
21 |
11 |
-1 |
0.747 |
0.902 |
-1.06 |
11.6 |
1.4 |
138 |
|
21 |
11 |
0.789 |
-1.06 |
0.933 |
11.6 |
1.4 |
138 |
22 |
11.5 |
-1 |
0.752 |
0.914 |
-1.04 |
10.7 |
1.14 |
133 |
|
22 |
11.5 |
0.779 |
-1.04 |
0.935 |
10.7 |
1.14 |
133 |
23 |
12 |
-1 |
0.743 |
0.924 |
-1.01 |
9.91 |
0.926 |
128 |
|
23 |
12 |
0.779 |
-1.02 |
0.947 |
9.91 |
0.926 |
128 |
Both algorithms try to find the lowest power whose R^2 is above the given threshold (0.90). If no power achieves the threshold then the threshold is decremented repeatedly by 0.05 until a solution is found. Since the Sage version finds a maximum of 0.887, it carries out the decrementing procedure. The UCLA version finds three values just above 0.90, and picks the lowest.
Modules
Unequal beta, module diff=44% |
identical beta, module diff=0.9% |
|---|---|
turquoise blue grey |
turquoise blue grey |
0 Comments