Table 5.
Outcomes of power analysis using merged SB1 and SB2/t-SB2 data to measure improvement before vs. after the transformation (Step 3).
| Merged untransformed (SB1 + SB2) |
Merged transformed (SB1 + t-SB2) |
Percent reduction |
|||||
|---|---|---|---|---|---|---|---|
| ANG SD | Effect size | Est. sample | ANG SD | Effect size | Est. sample | Est. sample (%) | |
| FTLD-Tau GM | 1.96 | 0.8 | 95 | 1.52 | 0.8 | 58 | –39 |
| (N = 20) | 1.96 | 0.5 | 242 | 1.52 | 0.5 | 146 | –40 |
| 1.96 | 0.2 | 1505 | 1.52 | 0.2 | 906 | –40 | |
| FTLD-Tau WM | 2.17 | 0.8 | 116 | 1.77 | 0.8 | 77 | –33 |
| (N = 20) | 2.17 | 0.5 | 296 | 1.77 | 0.5 | 197 | –34 |
| 2.17 | 0.2 | 1845 | 1.77 | 0.2 | 1224 | –34 | |
| FTLD-TDP GM | 1.35 | 0.8 | 45 | 1.20 | 0.8 | 36 | –20 |
| (N = 38) | 1.35 | 0.5 | 115 | 1.20 | 0.5 | 92 | –20 |
| 1.35 | 0.2 | 713 | 1.20 | 0.2 | 567 | –20 | |
| FTLD-TDP WM | 1.55 | 0.8 | 60 | 1.30 | 0.8 | 42 | –29 |
| (N = 38) | 1.55 | 0.5 | 152 | 1.30 | 0.5 | 107 | –30 |
| 1.55 | 0.2 | 943 | 1.30 | 0.2 | 661 | –30 | |
ANG, angular gyrus; Est., estimated; FTLD-Tau, frontotemporal lobar degeneration with inclusions of the tau protein; FTLD-TDP, frontotemporal lobar degeneration with inclusions of the transactive response DNA-binding protein 43 kDa; GM, gray matter; N, number of tissue samples; SB1, original staining batch; SB2, new staining batch (untransformed); SD, standard deviation; t-SB2, new staining batch (transformed); WM, white matter. Here, we show a power analysis to estimate the sample size necessary for an independent samples t-test testing pathology burden (i.e., mean ln %AO) in ANG against any other hypothetical brain region. Our aim was to measure the improvement in power after transformation by comparing (1) data merged from the original staining batch and the new staining batch without transformation (i.e., merged untransformed = SB1 + SB2), and (2) data merged from the original staining batch and the new staining batch after transformation (i.e., merged transformed = SB1 + t-SB2). We calculated ANG SD in these two sets of data and used it as an approximation of the overall variance (ANG vs. hypothetical region). We used effect sizes of 0.2, 0.5, and 0.8, corresponding to small, medium, and large effect sizes (Cohen, 1988), to estimate the sample size necessary (i.e., Est. sample) to detect a difference between mean ANG and another hypothetical regional mean. Each power analysis used alpha 0.05 and power 0.8. We measured the change between merged untransformed and merged transformed data by means of a percent reduction in estimated sample size.