Oiinkie
☆    

The Netherlands,
2013-01-16 17:42
(4484 d 18:27 ago)

Posting: # 9869
Views: 16,444
 

 bear for imbalanced data set [🇷 for BE/BA]

Post reply
Dear all,

As already indicated here in the past by Dr. Dan, when analyzing imbalanced datasets (in my case N=23, n1(RT)=12, n2(TR)=11, 2x2x2) with bear v2.5.3 the output of ANOVA_stat.txt and Statistical_summaries.txt do not allign. Different PE's and CI's are presented, while it should be the same in both files.

Thus, now it's the question which output is the correct one?

In my data set 23 subjects have completed the study (subject 24 is a dropout who only has data on T). When I actually include the data available of subject 24, bear crashes after NCA (which actually is logical)...

If you need more information or the actual data set (N=23) I am analyzing, please let me know.

Many thanks in advance!

Best regards,

Oiinkie
—
Regards,

Oiinkie
 
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2013-01-16 18:25
(4484 d 17:43 ago)

(edited on 2013-01-16 19:22)
@ Oiinkie
Posting: # 9871
Views: 15,126
 

 bear for imbalanced data set

Post reply
Dear Oiinkie,

Yes, please provide your data set if possible. Also if you have validated results with SAS or SPSS (as your post here) vs. bear with the same imbalanced data set, we would like to see that too. It will be helpful. Thank you so much.

❝ ...

❝ If you need more information or the actual data set (N=23) I am analyzing, please let me know.

❝ ...

—
All the best,
-- Yung-jin Lee
bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
 
Oiinkie
☆    

The Netherlands,
2013-01-16 20:18
(4484 d 15:50 ago)

@ yjlee168
Posting: # 9875
Views: 15,208
 

 bear for imbalanced data set

Post reply
Dear Yung-jin,

Thanks for your reply!

❝ Yes, please provide your data set if possible.


Sure, please find it below (csv). This is the full data set including subject 24 (the dropout who only completed period 1, T) which causes bear to crash after NCA. The imbalanced data set of concern is without this dropout, so N=23.

subject,sequence,period,time,conc
1,1,2,0.00,0
1,1,2,0.25,0
1,1,2,0.50,285
1,1,2,0.75,586
1,1,2,1.00,593
1,1,2,1.50,527
1,1,2,2.00,610
1,1,2,3.00,422
1,1,2,4.00,361
1,1,2,6.00,268
1,1,2,8.00,195
1,1,2,10.00,132
1,1,2,13.02,64.5
1,1,2,16.03,35.6
2,2,1,0.00,0
2,2,1,0.25,575
2,2,1,0.50,724
2,2,1,0.75,606
2,2,1,1.00,569
2,2,1,1.50,556
2,2,1,2.00,460
2,2,1,3.00,330
2,2,1,4.00,300
2,2,1,6.00,202
2,2,1,8.00,114
2,2,1,10.00,84.7
2,2,1,13.12,37.3
2,2,1,16.00,23
3,2,1,0.00,0
3,2,1,0.25,729
3,2,1,0.50,819
3,2,1,0.75,661
3,2,1,1.00,630
3,2,1,1.50,505
3,2,1,2.00,411
3,2,1,3.00,335
3,2,1,4.00,302
3,2,1,6.00,183
3,2,1,8.00,134
3,2,1,10.00,74.8
3,2,1,13.00,48.7
3,2,1,16.00,26.8
4,1,2,0.00,0
4,1,2,0.25,133
4,1,2,0.50,610
4,1,2,0.75,651
4,1,2,1.00,588
4,1,2,1.50,597
4,1,2,2.00,569
4,1,2,3.00,527
4,1,2,4.00,405
4,1,2,6.00,298
4,1,2,8.00,175
4,1,2,10.00,113
4,1,2,13.00,69
4,1,2,16.00,31.2
5,2,1,0.00,0
5,2,1,0.28,323
5,2,1,0.50,557
5,2,1,0.75,607
5,2,1,1.00,512
5,2,1,1.53,465
5,2,1,2.00,455
5,2,1,3.00,431
5,2,1,4.00,334
5,2,1,6.00,249
5,2,1,8.00,147
5,2,1,10.00,79.9
5,2,1,13.00,42.6
5,2,1,16.00,23.5
6,2,1,0.00,0
6,2,1,0.27,10.8
6,2,1,0.50,206
6,2,1,0.75,412
6,2,1,1.00,508
6,2,1,1.50,524
6,2,1,2.00,385
6,2,1,3.00,374
6,2,1,4.00,343
6,2,1,6.00,241
6,2,1,8.00,187
6,2,1,10.00,157
6,2,1,13.00,88.4
6,2,1,16.00,51.7
7,1,2,0.00,0
7,1,2,0.25,522
7,1,2,0.50,677
7,1,2,0.80,483
7,1,2,1.03,455
7,1,2,1.50,372
7,1,2,2.00,330
7,1,2,3.00,270
7,1,2,4.00,200
7,1,2,6.00,130
7,1,2,8.00,103
7,1,2,10.00,57.1
7,1,2,13.00,31.2
7,1,2,16.03,17.2
8,2,1,0.00,0
8,2,1,0.25,0
8,2,1,0.50,197
8,2,1,0.75,412
8,2,1,1.00,461
8,2,1,1.50,447
8,2,1,2.00,398
8,2,1,3.00,314
8,2,1,4.00,252
8,2,1,6.00,162
8,2,1,8.03,114
8,2,1,10.00,72.5
8,2,1,13.00,36.3
8,2,1,16.00,19
9,1,2,0.00,0
9,1,2,0.25,697
9,1,2,0.50,678
9,1,2,0.75,645
9,1,2,1.00,551
9,1,2,1.50,457
9,1,2,2.00,370
9,1,2,3.00,326
9,1,2,4.00,278
9,1,2,6.00,156
9,1,2,8.00,107
9,1,2,10.00,62
9,1,2,13.00,35.1
9,1,2,16.00,18.7
10,2,1,0.00,0
10,2,1,0.28,53.8
10,2,1,0.52,616
10,2,1,0.75,565
10,2,1,1.00,493
10,2,1,1.50,421
10,2,1,2.00,360
10,2,1,3.00,263
10,2,1,4.05,200
10,2,1,6.03,138
10,2,1,8.00,84.1
10,2,1,10.00,49
10,2,1,13.00,23.7
10,2,1,16.00,12
11,1,2,0.00,0
11,1,2,0.25,208
11,1,2,0.50,611
11,1,2,0.75,681
11,1,2,1.00,602
11,1,2,1.50,534
11,1,2,2.00,463
11,1,2,3.00,438
11,1,2,4.00,298
11,1,2,6.08,204
11,1,2,8.00,136
11,1,2,10.03,126
11,1,2,13.02,50.4
11,1,2,16.00,36.5
12,1,2,0.00,0
12,1,2,0.25,615
12,1,2,0.50,484
12,1,2,0.75,650
12,1,2,1.00,673
12,1,2,1.50,619
12,1,2,2.00,569
12,1,2,3.00,484
12,1,2,4.00,355
12,1,2,6.00,215
12,1,2,8.00,128
12,1,2,10.02,92.1
12,1,2,13.02,43.7
12,1,2,16.00,22.8
13,1,2,0.00,0
13,1,2,0.25,373
13,1,2,0.50,567
13,1,2,0.75,585
13,1,2,1.02,548
13,1,2,1.50,536
13,1,2,2.00,474
13,1,2,3.00,320
13,1,2,4.00,236
13,1,2,6.00,141
13,1,2,8.03,86.3
13,1,2,10.00,57.5
13,1,2,13.00,30.8
13,1,2,16.13,13.5
14,1,2,0.00,0
14,1,2,0.25,226
14,1,2,0.50,791
14,1,2,0.75,743
14,1,2,1.02,757
14,1,2,1.50,583
14,1,2,2.00,542
14,1,2,3.03,348
14,1,2,4.00,253
14,1,2,6.00,144
14,1,2,8.00,74.2
14,1,2,10.00,36.9
14,1,2,13.00,19.8
15,2,1,0.00,0
15,2,1,0.25,876
15,2,1,0.55,832
15,2,1,0.75,788
15,2,1,1.00,634
15,2,1,1.50,618
15,2,1,2.00,543
15,2,1,3.00,466
15,2,1,4.00,389
15,2,1,6.00,236
15,2,1,8.00,136
15,2,1,10.00,94.3
15,2,1,13.00,47.9
15,2,1,16.00,25.9
16,1,2,0.00,0
16,1,2,0.25,969
16,1,2,0.50,878
16,1,2,0.75,798
16,1,2,1.00,745
16,1,2,1.50,597
16,1,2,2.00,527
16,1,2,3.02,515
16,1,2,4.00,387
16,1,2,6.00,216
16,1,2,8.02,129
16,1,2,10.00,77.9
16,1,2,13.00,36.6
16,1,2,16.00,19
17,2,1,0.00,0
17,2,1,0.25,747
17,2,1,0.50,830
17,2,1,0.75,767
17,2,1,1.00,681
17,2,1,1.50,542
17,2,1,2.00,459
17,2,1,3.10,305
17,2,1,4.00,217
17,2,1,6.00,126
17,2,1,8.03,80
17,2,1,10.00,41.9
17,2,1,13.05,21.1
17,2,1,16.00,11.3
18,2,1,0.00,0
18,2,1,0.25,695
18,2,1,0.50,1102
18,2,1,0.77,983
18,2,1,1.00,874
18,2,1,1.50,698
18,2,1,2.00,633
18,2,1,3.00,482
18,2,1,4.00,359
18,2,1,6.00,211
18,2,1,8.00,120
18,2,1,10.00,63.9
18,2,1,13.00,27.9
18,2,1,16.00,13.6
19,1,2,0.00,0
19,1,2,0.25,377
19,1,2,0.50,753
19,1,2,0.75,689
19,1,2,1.02,663
19,1,2,1.53,586
19,1,2,2.00,497
19,1,2,3.03,402
19,1,2,4.03,329
19,1,2,6.03,216
19,1,2,8.02,133
19,1,2,10.02,76.4
19,1,2,13.02,38.7
19,1,2,16.00,16.7
20,2,1,0.00,0
20,2,1,0.25,371
20,2,1,0.50,859
20,2,1,0.75,735
20,2,1,1.00,665
20,2,1,1.50,549
20,2,1,2.00,507
20,2,1,3.00,334
20,2,1,4.00,279
20,2,1,6.00,149
20,2,1,8.00,75.3
20,2,1,10.00,36.5
20,2,1,13.00,13.6
21,1,2,0.00,0
21,1,2,0.25,149
21,1,2,0.50,996
21,1,2,0.75,858
21,1,2,1.00,977
21,1,2,1.50,749
21,1,2,2.00,641
21,1,2,3.02,498
21,1,2,4.00,368
21,1,2,6.00,248
21,1,2,8.00,131
21,1,2,10.00,76.7
21,1,2,13.00,40.4
21,1,2,16.00,18.5
22,1,2,0.00,0
22,1,2,0.28,0
22,1,2,0.50,136
22,1,2,0.75,324
22,1,2,1.02,505
22,1,2,1.52,726
22,1,2,2.00,742
22,1,2,3.00,690
22,1,2,4.00,587
22,1,2,6.00,383
22,1,2,8.00,230
22,1,2,10.02,95.8
22,1,2,13.00,40.4
22,1,2,16.00,17.4
23,2,1,0.00,0
23,2,1,0.25,46.6
23,2,1,0.50,688
23,2,1,0.75,833
23,2,1,1.00,796
23,2,1,1.50,836
23,2,1,2.00,744
23,2,1,3.00,617
23,2,1,4.00,484
23,2,1,6.00,309
23,2,1,8.00,202
23,2,1,10.00,137
23,2,1,13.00,72.4
23,2,1,16.00,35.5
24,2,1,0.00,0
24,2,1,0.25,689
24,2,1,0.50,949
24,2,1,0.75,775
24,2,1,1.00,778
24,2,1,1.50,679
24,2,1,2.00,606
24,2,1,3.00,434
24,2,1,4.00,346
24,2,1,6.00,200
24,2,1,8.00,118
24,2,1,10.00,65
24,2,1,13.00,33.1
24,2,1,16.00,15.2


Continued in next post...
—
Regards,

Oiinkie
 
Oiinkie
☆    

The Netherlands,
2013-01-16 20:19
(4484 d 15:50 ago)

@ yjlee168
Posting: # 9876
Views: 15,183
 

 bear for imbalanced data set

Post reply
1,1,1,0.00,0
1,1,1,0.25,0
1,1,1,0.50,73.8
1,1,1,0.75,248
1,1,1,1.00,404
1,1,1,1.50,460
1,1,1,2.00,456
1,1,1,3.00,423
1,1,1,4.00,368
1,1,1,6.00,218
1,1,1,8.00,167
1,1,1,10.00,103
1,1,1,13.00,61.9
1,1,1,16.00,31.8
2,2,2,0.00,0
2,2,2,0.32,628
2,2,2,0.50,701
2,2,2,0.75,596
2,2,2,1.00,638
2,2,2,1.50,532
2,2,2,2.00,475
2,2,2,3.00,353
2,2,2,4.00,292
2,2,2,6.00,190
2,2,2,8.02,135
2,2,2,10.00,70
2,2,2,13.03,32.9
2,2,2,16.00,14.4
3,2,2,0.00,0
3,2,2,0.25,158
3,2,2,0.50,649
3,2,2,0.75,652
3,2,2,1.00,600
3,2,2,1.50,515
3,2,2,2.00,425
3,2,2,3.00,392
3,2,2,4.00,285
3,2,2,6.00,183
3,2,2,8.00,146
3,2,2,10.00,87.7
3,2,2,13.00,50.7
3,2,2,16.00,31
4,1,1,0.00,0
4,1,1,0.25,219
4,1,1,0.50,668
4,1,1,0.75,661
4,1,1,1.00,704
4,1,1,1.50,665
4,1,1,2.00,647
4,1,1,3.00,525
4,1,1,4.00,479
4,1,1,6.00,384
4,1,1,8.00,230
4,1,1,10.00,156
4,1,1,13.00,83.1
4,1,1,16.00,46.3
5,2,2,0.00,0
5,2,2,0.25,316
5,2,2,0.50,629
5,2,2,0.75,558
5,2,2,1.00,512
5,2,2,1.50,421
5,2,2,2.00,370
5,2,2,3.00,312
5,2,2,4.00,258
5,2,2,6.00,167
5,2,2,8.00,106
5,2,2,10.00,61.9
5,2,2,13.00,31.5
5,2,2,16.00,17.1
6,2,2,0.00,0
6,2,2,0.25,12.6
6,2,2,0.50,152
6,2,2,0.75,335
6,2,2,1.00,444
6,2,2,1.50,551
6,2,2,2.00,428
6,2,2,3.00,365
6,2,2,4.00,317
6,2,2,6.02,257
6,2,2,8.00,182
6,2,2,10.00,182
6,2,2,13.00,110
6,2,2,16.00,61
7,1,1,0.00,0
7,1,1,0.25,663
7,1,1,0.50,600
7,1,1,0.75,532
7,1,1,1.00,422
7,1,1,1.50,402
7,1,1,2.00,341
7,1,1,3.00,273
7,1,1,4.00,240
7,1,1,6.00,169
7,1,1,8.00,106
7,1,1,10.00,69.5
7,1,1,13.00,53
7,1,1,16.00,26.9
8,2,2,0.00,0
8,2,2,0.25,43.9
8,2,2,0.50,418
8,2,2,0.75,424
8,2,2,1.00,406
8,2,2,1.50,444
8,2,2,2.00,381
8,2,2,3.00,303
8,2,2,4.00,247
8,2,2,6.00,161
8,2,2,8.02,97.9
8,2,2,10.03,65.8
8,2,2,13.03,29.3
8,2,2,16.00,16.2
9,1,1,0.00,0
9,1,1,0.27,486
9,1,1,0.50,567
9,1,1,0.75,542
9,1,1,1.00,467
9,1,1,1.50,459
9,1,1,2.00,356
9,1,1,3.00,317
9,1,1,4.00,263
9,1,1,6.00,184
9,1,1,8.00,101
9,1,1,10.02,62.9
9,1,1,13.00,29.2
9,1,1,16.00,14.9
10,2,2,0.00,0
10,2,2,0.25,181
10,2,2,0.50,553
10,2,2,0.75,534
10,2,2,1.00,506
10,2,2,1.50,449
10,2,2,2.00,389
10,2,2,3.00,304
10,2,2,4.00,242
10,2,2,6.00,176
10,2,2,8.00,117
10,2,2,10.00,69.4
10,2,2,13.00,29.1
10,2,2,16.00,17
11,1,1,0.00,0
11,1,1,0.25,79.3
11,1,1,0.50,422
11,1,1,0.75,651
11,1,1,1.00,602
11,1,1,1.50,545
11,1,1,2.00,458
11,1,1,3.00,348
11,1,1,4.05,316
11,1,1,6.03,223
11,1,1,8.00,144
11,1,1,10.00,113
11,1,1,13.00,49.1
11,1,1,16.00,27.6
12,1,1,0.00,0
12,1,1,0.25,386
12,1,1,0.50,685
12,1,1,0.75,517
12,1,1,1.00,523
12,1,1,1.50,566
12,1,1,2.00,422
12,1,1,3.00,388
12,1,1,4.03,321
12,1,1,6.00,178
12,1,1,8.00,110
12,1,1,10.00,79.7
12,1,1,13.00,32.2
12,1,1,16.00,17.8
13,1,1,0.00,0
13,1,1,0.50,675
13,1,1,0.75,526
13,1,1,1.00,540
13,1,1,1.50,410
13,1,1,2.00,384
13,1,1,3.00,281
13,1,1,4.00,229
13,1,1,6.00,114
13,1,1,8.03,74.2
13,1,1,10.00,49.8
13,1,1,13.00,26
13,1,1,16.00,11
14,1,1,0.00,0
14,1,1,0.32,547
14,1,1,0.60,798
14,1,1,0.88,629
14,1,1,1.00,573
14,1,1,1.50,569
14,1,1,2.00,473
14,1,1,3.00,327
14,1,1,4.00,256
14,1,1,6.05,144
14,1,1,8.00,77.6
14,1,1,10.00,38.5
14,1,1,13.00,17.5
15,2,2,0.00,0
15,2,2,0.25,0
15,2,2,0.50,480
15,2,2,0.75,735
15,2,2,1.00,673
15,2,2,1.50,614
15,2,2,2.00,539
15,2,2,3.00,462
15,2,2,4.00,335
15,2,2,6.00,227
15,2,2,8.00,121
15,2,2,10.00,76.5
15,2,2,13.00,35.6
15,2,2,16.00,19.2
16,1,1,0.00,0
16,1,1,0.25,827
16,1,1,0.50,728
16,1,1,0.75,851
16,1,1,1.00,729
16,1,1,1.50,597
16,1,1,2.00,489
16,1,1,3.00,445
16,1,1,4.00,359
16,1,1,6.00,165
16,1,1,8.00,111
16,1,1,10.00,62.1
16,1,1,13.02,27
16,1,1,16.00,14.7
17,2,2,0.00,0
17,2,2,0.25,825
17,2,2,0.50,960
17,2,2,0.75,850
17,2,2,1.00,748
17,2,2,1.50,646
17,2,2,2.00,536
17,2,2,3.00,378
17,2,2,4.00,303
17,2,2,6.00,189
17,2,2,8.00,124
17,2,2,10.02,67.1
17,2,2,13.00,29.7
17,2,2,16.00,16.1
18,2,2,0.00,0
18,2,2,0.25,125
18,2,2,0.50,963
18,2,2,0.75,929
18,2,2,1.02,929
18,2,2,1.52,707
18,2,2,2.00,639
18,2,2,3.00,553
18,2,2,4.00,400
18,2,2,6.00,259
18,2,2,8.02,135
18,2,2,10.02,86.4
18,2,2,13.00,37.1
18,2,2,16.00,16.5
19,1,1,0.00,0
19,1,1,0.25,614
19,1,1,0.50,768
19,1,1,0.75,668
19,1,1,1.00,582
19,1,1,1.50,562
19,1,1,2.00,476
19,1,1,3.00,373
19,1,1,4.00,301
19,1,1,6.00,179
19,1,1,8.00,114
19,1,1,10.00,73.9
19,1,1,13.05,37.1
19,1,1,16.00,20.9
20,2,2,0.00,0
20,2,2,0.25,220
20,2,2,0.50,423
20,2,2,0.78,602
20,2,2,1.02,637
20,2,2,1.53,541
20,2,2,2.00,592
20,2,2,3.02,439
20,2,2,4.03,330
20,2,2,6.00,203
20,2,2,8.02,90
20,2,2,10.00,44.8
20,2,2,13.00,20.4
21,1,1,0.00,0
21,1,1,0.25,493
21,1,1,0.53,1263
21,1,1,0.75,991
21,1,1,1.03,918
21,1,1,1.50,818
21,1,1,2.00,675
21,1,1,3.00,484
21,1,1,4.00,404
21,1,1,6.00,215
21,1,1,8.00,134
21,1,1,10.00,81.4
21,1,1,13.00,45.1
21,1,1,16.00,24.4
22,1,1,0.00,0
22,1,1,0.25,142
22,1,1,0.50,625
22,1,1,0.75,829
22,1,1,1.00,779
22,1,1,1.50,648
22,1,1,2.00,541
22,1,1,3.00,515
22,1,1,4.00,375
22,1,1,6.00,253
22,1,1,8.00,132
22,1,1,10.00,65.9
22,1,1,13.00,24.3
22,1,1,16.00,13.8
23,2,2,0.00,0
23,2,2,0.27,48.8
23,2,2,0.50,561
23,2,2,0.75,616
23,2,2,1.02,701
23,2,2,1.50,691
23,2,2,2.00,569
23,2,2,3.03,551
23,2,2,4.00,428
23,2,2,6.00,257
23,2,2,8.00,186
23,2,2,10.00,91.8
23,2,2,13.00,48.9
23,2,2,16.00,19.3


❝ Also if you have validated results with SAS or SPSS (as your post here) vs. bear with the same imbalanced data set, we would like to see that too.


Unfortunately, I currently do not have access to SAS or SPSS (dead laptop...) so I was not able to validate results. This is actually the reason why I asked which output file of bear "tells the truth": ANOVA_stat.txt or Statistical_summaries.txt? Would you already be able to answer this question? :confused:

Many thanks!

Oiinkie
—
Regards,

Oiinkie
 
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2013-01-16 20:43
(4484 d 15:25 ago)

@ Oiinkie
Posting: # 9879
Views: 15,038
 

 bear for imbalanced data set

Post reply
Dear Oiinkie,

Thank you for your data set. At the end of 2010, I tried to fix the inconsistency between two outputs. Then I faced the same problem as you did: which one should be correct? Thus why I did not release bear v2.5.4 to public. The data set I used to test bear was obtained from full data set by deleting one subject's data. I was not sure if I was doing the right thing at that moment. I thought that I should confirm the final results with SAS. Don't quite remember everything right now. Anyway, I will test bear again with your data set, as well as using SAS. Sorry about this.

❝ ...

❝ Unfortunately, I currently do not have access to SAS or SPSS (dead laptop...) so I was not able to validate results. This is actually the reason why I asked which output file of bear "tells the truth": ANOVA_stat.txt or Statistical_summaries.txt? Would you already be able to answer this question? :confused:

❝ ...

—
All the best,
-- Yung-jin Lee
bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
 
ElMaestro
★★★

Denmark,
2013-01-16 23:10
(4484 d 12:58 ago)

@ yjlee168
Posting: # 9880
Views: 15,016
 

 bear for imbalanced data set

Post reply
Dear yjlee,

❝ Thank you for your data set. At the end of 2010, I tried to fix the inconsistency between two outputs. Then I faced the same problem as you did: which one should be correct?


You can easily check the result by fitting an lm without intertcept, e.g. with Formulation specified as the first term. The difference in 'LSMeans' for lack of better wording on the log scale will be directly extractable from the coefficients vector's first and second values.
—
Pass or fail!
ElMaestro
 
Oiinkie
☆    

The Netherlands,
2013-01-22 14:28
(4478 d 21:41 ago)

@ yjlee168
Posting: # 9897
Views: 14,571
 

 bear for imbalanced data set

Post reply
Dear Yung-jin (and other forum members),

❝ I will test bear again with your data set, as well as using SAS.


Have you already been able to test the data set in order to find out whether ANOVA_stat.txt or Statistical_summaries.txt gives the "correct answer"? This would help me a lot.

Many thanks!
—
Regards,

Oiinkie
 
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2013-01-22 16:06
(4478 d 20:03 ago)

@ Oiinkie
Posting: # 9898
Views: 14,668
 

 comparison with PHX

Post reply
Dear Oiinkie,

maybe you are interested in an independent evaluation (Phoenix 6.3). Only Cmax since I don’t know your AUC-algo.
  • subject(sequence) random, n=24/23:
    104.98% (99.83–110.40%), CVintra 9.96%
  • subject(sequence) fixed, n=24/23:
    104.76% (99.61–110.18%), CVintra 9.95%
  • subject(sequence) random, n=23/23:
    104.76% (99.61–110.18%) CVintra 9.95%
  • subject(sequence) fixed, n=23/23:
    104.76% (99.61–110.18%) CVintra 9.95%
EMA wants only the last variant.
—
Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
 
Oiinkie
☆    

The Netherlands,
2013-01-23 13:58
(4477 d 22:10 ago)

@ Helmut
Posting: # 9904
Views: 14,570
 

 comparison with PHX

Post reply
Dear Helmut,

Many thanks for your assessment!

❝ • subject(sequence) fixed, n=23/23:

❝ 104.76% (99.61–110.18%) CVintra 9.95%

❝ EMA wants only the last variant.


I am aware of these "rules" of the EMA ;-) (fixed rather dodgy); I also ran the analysis on n=23/23. The PE, 90% CI and CVintra mentioned in ANOVA_stat.txt are exactly the same! Statistical_summaries.txt gives 104.573% (99.430-109.981%). It seems that with transfer of the results from ANOVA_stat.txt to Statistical_summaries.txt (I assume the latter is compiled based on the former) something goes wrong for imbalanced data sets...

❝ Only Cmax since I don’t know your AUC-algo.


AUC would be AUC(0-t) calculated by linear trapezoid method (bear's standard; not the nicest one, but I want to stick with method of the CRO for this reanalysis of this very old study). AUC(0-inf) is not critical at the moment (secondary parameter).
—
Regards,

Oiinkie
 
Oiinkie
☆    

The Netherlands,
2013-01-23 16:07
(4477 d 20:02 ago)

@ Oiinkie
Posting: # 9906
Views: 14,489
 

 comparison with PHX

Post reply
In addition to the previous post and for information purposes...

❝ It seems that with transfer of the results from ANOVA_stat.txt to Statistical_summaries.txt (I assume the latter is compiled based on the former) something goes wrong for imbalanced data sets...


After looking into the output files a bit more, Statistical_summaries.txt presents the PE by calculating (geometric mean T)/(geometric mean R), which is in general not correct for imbalanced data sets (PE should be the square root of (lower limit CI)*(upper limit CI)). It looks like (wild guess :confused:) that for Statistical_summaries.txt first an incorrect PE is calculated and subsequently its 90% CI by using the MSE. This all would lead to an incorrect PE and 90% CI in Statistical_summaries.txt.

Could my assessment be correct?
—
Regards,

Oiinkie
 
yicaoting
★    

NanKing, China,
2013-02-15 19:58
(4454 d 16:11 ago)

@ Helmut
Posting: # 10037
Views: 14,299
 

 comparison with SAS

Post reply

❝ • subject(sequence) random, n=24/23:

❝ 104.98% (99.83–110.40%), CVintra 9.96%

❝ • subject(sequence) fixed, n=24/23:

❝ 104.76% (99.61–110.18%), CVintra 9.95%

❝ • subject(sequence) random, n=23/23:

❝ 104.76% (99.61–110.18%) CVintra 9.95%

❝ • subject(sequence) fixed, n=23/23:

❝ 104.76% (99.61–110.18%) CVintra 9.95%[/list]


Using the same data with SAS's GLM procedure, results are:
n=24/23
        Least Squares Means for Effect formulation

               Difference
                  Between    90% Confidence Limits for
   i    j           Means       LSMean(i)-LSMean(j)

   1    2        0.046485     -0.003939     0.096910


n=23/23
        Least Squares Means for Effect formulation

               Difference
                  Between    90% Confidence Limits for
   i    j           Means       LSMean(i)-LSMean(j)

   1    2        0.046485       -0.003939     0.096910

Exactly the same. Transform the above results to PE and CI:
PE: 1.047582365
CI: 0.996068748 to 1.101761211
 
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2013-01-22 20:57
(4478 d 15:12 ago)

@ Oiinkie
Posting: # 9899
Views: 14,601
 

 bear for imbalanced data set

Post reply
Dear Oiinkie,

Not yet and still working on it. I will post the final results here. Apparently, validating with SAS does not provide enough information. Will figure out Elmaestro's method posted in this thread.

❝ Have you already been able to test the data set in order to find out whether ANOVA_stat.txt or Statistical_summaries.txt gives the "correct answer"? This would help me a lot.

—
All the best,
-- Yung-jin Lee
bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
 
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2013-03-01 02:01
(4441 d 10:08 ago)

@ Oiinkie
Posting: # 10137
Views: 14,098
 

 ANOVA_stat.txt is correct. → bear for imbalanced data set

Post reply
Dear Oiinkie, Helmut and yicaoting

Thank you all for running WNL (by Helmut) and SAS (by yicaoting) for imbalanced data provided by Oiinkie. This is definitely a great Forum.
O.K., bear will crash with incomplete data. The original question was: with imbalanced data, bear got two output files (ANOVA_stat.txt and Statistical_summaries.txt). The results showed inconsistent. After running the imbalance data, it can be concluded:
  1. the result of ANOVA_stat.txt is correct based on WNL & SAS runs. Oiinkie probably has already known that.
  2. the PE from Statistical_summaries.txt is also incorrect as pointed by Oiinkie.
  3. the error only occurs in imbalanced data; for incomplete data, bear will be crashed.
I will try to fix these errors/bugs asap. Finally, I am very sorry to keep you all waiting, as well as any inconveniences due to these errors/bugs.
I only abstract the differences as mentioned by Oiinkie as follows:

[from ANOVA_stat.txt] ...
*** Classical (Shortest) 90% C.I. for lnCmax ***

        Point Estimate   CI90 lower   CI90 upper
Ratio          104.758       99.607      110.176

---------------------- Two One-Sided Tests (TOST) -------------------------

     TOST   T value   P value
1 T_lower    -9.141     0.000
2 T_upper    -6.089     0.000

**Interpretation:
Ho: Theta <  0.80000  or  Theta >  1.25000
Ha:  0.80000  < or = Theta < or =  1.25000
where Theta = Mean_Test/Mean_Ref.
Because all P values are less than 0.05, we will reject the null hypothesis (Ho).
BE acceptance criterion is set within the range of  80.000  -  125.000 % .
...
Intra_subj. CV = 9.953 %
...

and
[from Statistical_summaries.txt] ...
Statistical Summaries for Pivotal Parameters of Bioequivalence (N= 23 )
(cont'd)
-------------------------------------------------------------------------

    Parameters  F values  P vales    PE (%)  lower 90%CI   upper 90%CI
1         Cmax     1.643    0.214         -            -            -
2       AUC0-t     1.466    0.239         -            -            -
3     AUC0-inf     1.286    0.270         -            -            -
4     ln(Cmax)     2.516    0.128   104.573       99.4304      109.981  ← Error!
5   ln(AUC0-t)     1.330    0.262   103.317       98.5645      108.299
6 ln(AUC0-inf)     1.173    0.291   103.239       98.2956      108.433
...
ps. please ignore AUC stuffs. I just arbitrarily put 100 as the the given dose in order to complete the run with bear.

ps2: A good news is that I think the inconsistency has been fixed back to 2011-12-22 (bear v2.5.4, not released to the public yet). However, the crash with incomplete data has been not yet.[edited]

❝ Have you already been able to test the data set in order to find out whether ANOVA_stat.txt or Statistical_summaries.txt gives the "correct answer"? This would help me a lot.

—
All the best,
-- Yung-jin Lee
bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
 
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2013-03-15 08:58
(4427 d 03:11 ago)

@ yjlee168
Posting: # 10198
Views: 14,114
 

 bear for incomplete data set?

Post reply
Dear all,

I am working on bear v2.5.4 right now and try to solve this problem. The error (something like ..error in data.frame... arguments imply differing number of rows...) with incomplete dataset with bear is due to "...data.frame [in R] is an object, which resembles a table from Excel, it has (must have!) the same number of rows in each column." That's funny rule in R.:confused: It's not like Excel at all. Apparently, R treats data.frame() pretty much like matrix, I guess. Thus, it won't allow not equal number of rows in column. Since index of subject number is different with incomplete dataset, we need to restructure all data.frame in bear using plyr (split()) or cbind(), etc.. Just leave a note here for later reference. Probably I will leave as it was for v2.5.4 if I cannot solve it within one or two days.

❝ ...

❝ 3. the error only occurs in imbalanced data; for incomplete data, bear will be crashed.

❝ ...

—
All the best,
-- Yung-jin Lee
bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
 
yicaoting
★    

NanKing, China,
2013-02-15 19:48
(4454 d 16:21 ago)

@ Oiinkie
Posting: # 10036
Views: 14,364
 

 imbalanced or incomplete data

Post reply
Dear Oiinkie, yjlee168, ElMaestro, Helmut,

I don't think Oiinkie's data is a set of imbalanced data, but a set of incomplete data.

In my personal opinion, In 2*2*2 BE study,
imbalanced data: the number of subjects of Sequence(RT) is not equal to that of Sequence(TR).

incomplete data: the number of subjects in Period 1 is not equal to that in Period 2.

Oiinkie's data included Subject(24) in Period 1 and missed Subject(24) in Period 2, so Oiinkie's data is a set of incomplete data. If we delete Subject(24) in Period 1, it becomes a set of imbalanced data.

Is it right?
 
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2013-02-16 15:41
(4453 d 20:28 ago)

@ yicaoting
Posting: # 10043
Views: 14,237
 

 incomplete data

Post reply
Dear Zhang Yong!

❝ imbalanced data: the number of subjects of Sequence(RT) is not equal to that of Sequence(TR).

❝ incomplete data: the number of subjects in Period 1 is not equal to that in Period 2.

❝

❝ Is it right?


Exactly.
—
Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
 
UA Flag
Activity
 Admin contact
23,424 posts in 4,927 threads, 1,672 registered users;
79 visitors (0 registered, 79 guests [including 3 identified bots]).
Forum time: 13:09 CEST (Europe/Vienna)

There are two possible outcomes: if the result confirms the
hypothesis, then you’ve made a measurement. If the result is
contrary to the hypothesis, then you’ve made a discovery.    Enrico Fermi

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5