diff --git a/doc/Documentation.pdf b/doc/Documentation.pdf index 6507de1962cf0e6691a7cadaee7c8bf07999f9a6..e412a80d343be1e8e06f9ca1362af6d9d3058c22 100644 Binary files a/doc/Documentation.pdf and b/doc/Documentation.pdf differ diff --git a/src/R/7.summaryFinal.R b/src/R/7.summaryFinal.R index 1779e39b4a5406ab7047fce8767deece5bc51764..93af655e67024ae24f28c52ab80dbc94bb2b3bea 100755 --- a/src/R/7.summaryFinal.R +++ b/src/R/7.summaryFinal.R @@ -30,8 +30,7 @@ par.l = list() # Hard-coded parameters par.l$verbose = TRUE par.l$FDR_threshold = c(0.001, 0.01, 0.05,0.1,0.2) -par.l$probsThreshold = c(0.01, 0.99) -par.l$probsThreshold = c(0.05, 0.95) +par.l$probsThreshold = c(0, 1) par.l$expressionThreshold = 2 par.l$cohensDThreshold = 0.1 par.l$classes_CohensD = c("small", "medium", "large", "very large") @@ -264,21 +263,6 @@ CME.delta = estimates[2] # as seen in Figure 6.1a, can derail central matching. The MLE method is more stable, but pays the price of possibly increased bias. -# 2. Should the variance associated with this weighted.mean T stat value be based on the variance of the T stat scores, the variance of the mean, or both? - -# 3. Why not estimate the SD of the distribution directly, without the need to provide manual variance estimations? Because then each weigted mean value -# is treated as if it came from the same population? - -# 4. library(Hmisc) -> wtd.var(x, weights) - -# 5 . Variance estimation is based on before the centralization, is this correct? Yes, independent - -# 6. Is centralization actually correct at all since the scores come from a different population? Yes, if assumed they come from a normal distribution - -# NEW: Estimate variance of T score estimate with Welch corrected df (test$parameter) - -# READ http://genomicsclass.github.io/book/pages/t-tests_in_practice.html - output.global.TFs$weighted_Tstat_centralized = output.global.TFs$weighted_Tstat - MLE.delta output.global.TFs$variance = as.numeric(output.global.TFs$variance)