Varitation in drug admistration parameter

[dancoster]

1. Explore what what is the variation of administration type (PO, IV, drips etc.) of each drug. Is there any variation?
2. Explore what what is the variation of dosage? Is there any variation?

Box Plot (X-Axis: PO, IV, Drips, Y-axis: Ratio, P-value per each administration types).

[dancoster]

[1] Per each sig. pair of varied-admitted drug, please show the effect of different admistarition.

[dancoster]

[1] Boxplot figure (specific pair): (A) flip axis (B) per each group write size (n) and p-value (C) and ration (D) add title+parametrs of function (before = 4h, after=6h)

[rfeinsod]

We want to generate adataframe that the first column is a pair, second administration type, third - num of instances, fourth - pvalue of the pair per administration type.

[dancoster]

There is a bug with the p-values (all are NaN or 1). Additionally, please notice that it will be easier for us to analyze it with a data frame (and not multiple plots). Can you please generate this dataframe?

[rfeinsod]

plots_different_admin.zip

[dancoster]

Thx, how did you select these pairs?

[rfeinsod]

These are all the pairs with a Bonferroni corrected pvalue under 0.05

[dancoster]

mmm I think there is a mistake here, some known pairs are misiing

[rfeinsod]

Do you have a specific example?

[rfeinsod]

Okay, all missing pairs are pairs where there is only one administration type. I can generate plots for them but the plots wouldn't show anything meaningful.

[dancoster]

ok, let's try to generate similar boxplot per each significant pair, but now stratify on the dosage of the administered drug.

[dancoster]

• Understand diff between AMOUNT and TOTALAMOUNT
• Stratify to K groups, where each group k has at least N samples.

[rfeinsod]

Dosage stratified.zip

[dancoster]

• Title Bon Corrected P-value of all samples
• Add number of samples per dosage + raw p-value
• Add boxplot of all samples and add x-tick the raw p-value the before the Bonferroni correction.
• Per each group - calculate the non-corrected
• Filter samples with at least 100 samples (?) -same value as in the config of the p-value generation
• Do the same for administration type.
• Present few pairs (with known clinical relevancy) and try come with recommendation for modeling.

[rfeinsod]

• [ V ] Title Bon Corrected P-value of all samples
• [ V ] Add number of samples per dosage + raw p-value
• [ V] Add boxplot of all samples and add x-tick the raw p-value the before the Bonferroni correction.
• [ ] Per each group - calculate the bon-corrected
• [V ] Filter samples with at least 100 samples (?) -same value as in the config of the p-value generation
• [ ] Do the same for administration type.
• [ ] Present few pairs (with known clinical relevancy) and try come with recommendation for modeling.

Will continue working on the rest

[dancoster]

Looks better, a few minor comments:

• Take only 3 digits after dot in each p-value
• Put all in different color
• Add to the X-axis tick the Mean-+ STD of the ratio of each group
• Why does all p-value and the p-value in the title are different?
• what's the name of algorithm that you were used to stratify the groups?
• add also units to the title
• Add table per pair with the percentage of patients that were removed because of units inalignment.
• Add p-values that compare the distribution between boxplots. Please add some more examples

[rfeinsod]

The Pvalue of the title is after the Bonferroni correction. I'm adding some examples now and will re-send all soon after additional changes. To stratify the bins I used pd.qcut to create quantile bins and then split the dataframe accordingly

[rfeinsod]

Take only 3 digits after dot in each p-value -Done Put all in different color - Done Add to the X-axis tick the Mean-+ STD of the ratio of each group - Done Why does all p-value and the p-value in the title are different? - Explained what's the name of algorithm that you were used to stratify the groups? - Explained add also units to the title - Done Add table per pair with the percentage of patients that were removed because of units inalignment. - Done Add p-values that compare the distribution between boxplots. - Done, they don't look, will fix Please add some more examples - Done

[dancoster]

[1] Do you use the full dataset here? (or only 10,000 patients) [2] Please extract only box plots with sig. difference [3] Please make one color for 'all', and the same color to all the groups [4] Changetd to: [5] Please extract a table with all the pairs which contain patients that were removed due to unit mismatch, and per each pair report the percentage [6] Please extract examples with sig. p-value between boxplots (like piggyback<>Potassium) What did you mean by "Add p-values that compare the distribution between boxplots. - Done, they don't look, will fix" [7] Please generate the similar plot but now stratify of administration type

[dancoster]

• Think how to compare GAM function

[rfeinsod]

1) The Full dataset 2) Ok, Examples Attached 3) OK 4) OK 5)Attached

removals.csv

[dancoster]

[5]Adjust 'removals.csv' -(a) calculate p-value per each pair on all samples, (b) calc p-value only of samples with the major type of units (c) add column if there are multiple admistration types per this drug (1/0) (d) calculate the f0ld-change (ratio) between the two types of p-value. [8] Fix title issue and upload updated tables.