英文:
Python: creating plot based on observation dates (not as a time series)
问题
我有以下的数据集
dfid medication_date1 2000-01-011 2000-01-041 2000-01-062 2000-04-012 2000-04-022 2000-04-03
我想首先将数据集重塑为每位患者首次观察后的天数:
id day1 day2 day3 day41 yes no no yes2 yes yes yes no
最终要创建一个使用上述表格的绘图: 列是日期，如果是“yes”则为黑色，如果不是则为白色。
非常感谢任何帮助。
英文:
I have the following dataset
dfid medication_date1 2000-01-011 2000-01-041 2000-01-062 2000-04-012 2000-04-022 2000-04-03
I would like to first reshape the data set into days after the first observation per patient:
id day1 day2 day3 day41 yes no no yes2 yes yes yes no
in order to ultimately create a plot with the above table: columns the dates and in black if yes, and white if not.
any help really appreciated it
答案1
得分: 2
将稀疏的Series（'yes'药物）转换为稠密的Series，通过添加缺失的日期（'no'药物），然后重置Series的索引（2000-01-01 -> 0, 2000-04-01 -> 0）。最后，重新塑造您的数据框。
def f(sr):# 创建缺失的日期dti = pd.date_range(sr.min(), sr.max(), freq='D')# 用'yes'或'no'填充Seriesreturn (pd.Series('yes', index=sr.tolist()).reindex(dti, fill_value='no').reset_index(drop=True))df['medication_date'] = pd.to_datetime(df['medication_date'])out = (df.groupby('id')['medication_date'].apply(f).unstack(fill_value='no').rename(columns=lambda x: f'day{x+1}').reset_index())
输出:
>>> outid day1 day2 day3 day4 day5 day60 1 yes no no yes no yes1 2 yes yes yes no no no
更新
import matplotlib.pyplot as pltfrom matplotlib.colors import LinearSegmentedColormapcolors = ["white", "black"]cmap = LinearSegmentedColormap.from_list('Custom', colors, len(colors))plt.matshow(out.set_index('id').eq('yes').astype(int), cmap=cmap)plt.show()
英文:
Transform the sparse Series ('yes' medication) to dense Series by adding missing days ('no' medication) then reset the Series index (2000-01-01 -> 0, 2000-04-01 -> 0). Finally, reshape your dataframe.
def f(sr):# Create missing datesdti = pd.date_range(sr.min(), sr.max(), freq='D')# Fill the Series with 'yes' or 'no'return (pd.Series('yes', index=sr.tolist()).reindex(dti, fill_value='no').reset_index(drop=True))df['medication_date'] = pd.to_datetime(df['medication_date'])out = (df.groupby('id')['medication_date'].apply(f).unstack(fill_value='no').rename(columns=lambda x: f'day{x+1}').reset_index())
Output:
>>> outid day1 day2 day3 day4 day5 day60 1 yes no no yes no yes1 2 yes yes yes no no no
Update
import matplotlib.pyplot as pltfrom matplotlib.colors import LinearSegmentedColormapcolors = ["white", "black"]cmap = LinearSegmentedColormap.from_list('Custom', colors, len(colors))plt.matshow(out.set_index('id').eq('yes').astype(int), cmap=cmap)plt.show()
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论