View source: R/normalization.R
mu_calculate_threshold | R Documentation |
Calculate a reasonable threshold value to replace zeros
mu_calculate_threshold(intensity_df, value_transform = 1/2)
intensity_df |
intensity values across many samples |
value_transform |
how much to scale the value (default = 1/2) |
Not reported or missing values in our data cause all kinds of problems, and for data with proportional error, log-transforms mean we can't just set them to 0 either. In addition, just setting to zero inflates the differences of values. So, a reasonable value for noise is 1/2 of the lowest observed value in a *normal* like distribution. To achieve that for data with proportional data, we do a log-transform first. Because we want to use *all* the data across samples, we might have some weird outliers too. So we don't use the values directly from the distrubtion, but use 'boxplot.stats' to get a reasonable handle on the distribution as well, and take a fraction of the lowest value in the distribution description.
double
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.