Sample size calculation #519
Conversation
…n significance from t-test
…e calculation function
# Conflicts: # protzilla/methods/data_analysis.py # ui/runs/form_mapping.py # ui/runs/forms/data_analysis.py
# Conflicts: # protzilla/methods/data_analysis.py # ui/runs/forms/data_analysis.py
| if intensity_name is None: | ||
| intensity_name = "Normalised iBAQ" |
There was a problem hiding this comment.
Assumes that data has to be normalized before feeding into the step. Otherwise the column doesn't exist. I would say that is an unnecessary limitation that is not transparent to the user
| if intensity_name is None: | ||
| intensity_name = "Normalised iBAQ" |
There was a problem hiding this comment.
I feel that this could just the default argument if it is set anyways. Or is there a reason why the default has to be None?
| self.fields["t_test_results"].choices = get_t_test_results(run) | ||
|
|
||
| class PowerAnalysisSampleSizeCalculationForm(MethodForm): |
There was a problem hiding this comment.
In general, you should run a code formatter. Often there too few blank lines (between methods or here above the class, you should leave two empty lines) and unnecessary whitespaces around equal signs (as below in the ...Field()
…ary-function of Sample Size Calculation have the same result
| intensity_name=intensity_name, | ||
| ) | ||
| sample_size = differentially_expressed_proteins_df.groupby('Group')['Sample'].count() | ||
| z_beta = fc_threshold * np.sqrt(sample_size/(2*variance_protein_group**2))-z_alpha |
There was a problem hiding this comment.
| z_beta = fc_threshold * np.sqrt(sample_size/(2*variance_protein_group**2))-z_alpha | |
| z_beta = fc_threshold * np.sqrt(sample_size / (2 * variance_protein_group)) - z_alpha |
I think the square is too much since we are already dealing with variances and not standard deviations. (Also some minor formatting issues)
…ved validation methods to separate file
…roteins_only, intensity_name)
…iduals. If so, the mean values per individual are used to calculate the power and sample size.
…ecked values from paper of Cairns
… the distribution in a violin plot
…n't be displayed anymore (WIP...)
| fig = go.Figure() | ||
|
|
||
| fig.add_trace( | ||
| go.Violin( | ||
| x=["Protein Groups"] * len(required_sample_sizes), | ||
| y=required_sample_sizes, | ||
| line_color=colors[1], | ||
| **violin_plot_args | ||
| ) | ||
| ) | ||
| fig.update_layout( |
There was a problem hiding this comment.
If you don't add traces to a figure dynamically (e.g. in a for loop), you can also pass the trace directly to go.Figure()
| violin_plot_args = dict( | ||
| meanline_visible=True, | ||
| box_visible=True, | ||
| scalemode='width', | ||
| spanmode='hard', | ||
| span=[0, required_sample_size_for_all_proteins], | ||
| fillcolor='rgba(0,0,0,0)' | ||
| ) |
There was a problem hiding this comment.
would pass these arguments directly to go.Violin(). Since you are not reusing them somewhere else, it just makes the code harder to read because these args are in a different place
|
|
||
| fig.add_trace( | ||
| go.Violin( | ||
| x=["Protein Groups"] * len(required_sample_sizes), |
There was a problem hiding this comment.
Would omit the x parameter and just use name="Protein Groups". It's easier to read
…thods "...for All Proteins"
meta file that includes an additional column that identifies the individual sample IDs.
Description
Implementation of sample_size_calculation to calculate the required sample size for a selected protein. For this purpose, the variance method was added and an output field was implemented to display the result. The variance and the sample size calculation-method are tested with test-data in test_power_analysis.py
Changes
sample_size_calculation + variance method:
Display output field:
Test:
Mergeability
blackCode review