Estimate Confidence

The get_confidence function allows you to easily estimate the confidence intervals for either growth parameters or the predicted growth curves.


Estimating confidence intervals for growth parameters


AMiGA can estimate mean and standard deviation of growth parameters if it fit growth curves based on multiple replicates. See How to infer summary statistics for pooled replicates. For example,


python amiga.py fit -i /Users/firasmidani/experiment/ -o "pooled_analysis" --pool-by "Isolate,Substrate" --sample-posterior


The above command will generate a summary file summary/pooled_analysis_summary.txt that will include the estiamted mean and standard deviation for a variety of growth parameters. If you would like to estimate the confidence intervals for these parameters, you can do the following


python amiga.py get_confidence -i /Users/firasmidani/experiment/summary/summary/pooled_analysis_summary.txt --type 'Parameters' --confidence 95


This will generate a new file summary/pooled_analysis_summary_confidence.txt where it will include also the lower and upper bounds for the 95% confidence interval of all growth parameters.


Estimating confidence intervals for growth curves

AMiGA can pool replicate curve and model them jointly. If requested by the user (--save-gp-data, it will save the predicted mean and covariance for the growth curves as well as the estimated Gaussian noise. For example,


python amiga.py fit -i /Users/firasmidani/experiment/ -o "pooled_analysis" --pool-by "Isolate,Substrate" --sample-posterior --save-gp-data


The above command will generate a text fiel derived/pooled_analysis_gp_data.txt which will have columns for the sample’s meta-data, in addition to:

  • mu: mean of the growth function per time point
  • Sigma: variance of the growth function per time point
  • mu1: mean of the growth rate function per time point
  • Sigma1: variance of the growth function per time point
  • Noise: measurement Noise (time-independent by default but time-dependent if you also use the --fix-noise argument).


You can estimate the confidence intervals for the growth function and growth rate function as follows:


python amiga.py get_confidence -i /Users/firasmidani/experiment/derived/pooled_analysis_gp_data.txt --type 'Curves' --confidence 95


This will generate a new file derived/pooled_analysis_gp_data_confidence.txt. This copy of the input file will have four additional columns for the lower (Low) and upper (Upper) confidence intervals of the growth function, and the lower (Low1) and upper (Upper1) confidence intervals of the growth rate function

By default, get_confidence will compute the confidence intervals without including sampling uncertainty (i.e. measurement noise). If you would like to include noise in the confidence interval, you must pass --include-noise.


Command-line arguments

To see the full list of arguments that amiga compare will accept, run

python compare.py --help

which will return the following message

usage: amiga.py [-h] -i INPUT --type {Parameters,Curves}
                [--confidence CONFIDENCE] [--include-noise] [--over-write]
                [--verbose]

Compute confidence intervals for parameters or curves.

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
  --type {Parameters,Curves}
  --confidence CONFIDENCE
                        Must be between 80 and 100. Default is 95.
  --include-noise       Include the estimated measurement noise when computing
                        confidence interval (For Curves Only).
  --over-write          Over-write file otherwise a new copy is made with
                        "_confidence" suffix
  --verbose


See more details for these arguments in Command Line Interface