How to display a bar category on the axis when the value is 0-CodePudding

I have a dataframe column with the category of the corresponding row blood pressure systolic and diastolic values as obtained by the following function:

def classify_bp(row):
    if row.SYS < 120 and row.DIA < 80:
        return "normal"
    elif (row.SYS >= 120 and row.SYS <= 129) and row.DIA < 80:
        return "elevated"
    elif (row.SYS >= 130 and row.SYS <= 139) or (row.DIA >= 80 and row.DIA <= 89):
        return "stage1"
    elif (row.SYS >= 140 and row.SYS <= 179) or (row.DIA >= 90 and row.DIA <= 119):
        return "stage2"
    elif row.SYS > 180 or row.DIA > 120:
        return "crisis"

If I apply this function to the blood pressure readings dataframe I can generate a new df column called "bp_class" with the proper blood pressure reading categorization for each row.

If I apply the value_counts() function to that column I obtain the following:

df_simple["bp_class"].value_counts(sort=False)
normal      37
stage1      34
elevated    15

I would like to produce a seaborn barchart with the 5 categories on the x axis, possibly always in the same order of the function (normal, elevated,stage1,stage2,crisis) and the frequencies for each including the categories with zero observations (which are not known to the value_counts() function).

CodePudding user response：

You can

CodePudding user response：

As an alternative, if you have a categorical dtype seaborn will extend the categorical axis without further action. Here using an index to assign index-data relation but then resetting the index to get the data in long format suitable for seaborn plots.

class_cat = pd.CategoricalDtype(['normal', 'elevated', 'stage1', 'stage2', 'crisis'], ordered=True)
df = pd.DataFrame(index=pd.Index(['normal', 'stage1', 'elevated'], dtype=class_cat),
                  data={"bp_class": [37, 34, 15]}).reset_index()

sns.barplot(data=df, x="index", y="bp_class")

This as sidenote to Stef's answer.

If you start from

counts = df_simple["bp_class"].value_counts(sort=False).to_frame()

Then we'd transform the index from there instead:

df = counts.set_axis(counts.index.astype(class_cat)).reset_index()