Figure Friday 2024 - week 51

Dash Python
,
23

For week 51, I focused on percentiles of required years experience covering various occupations.
Included are data where the CATEGORY column is ‘Prior work experience’, and where the ESTIMATE_TEXT columns contain the word ‘percentile’.
A PERCENTILE column was created by parsing text in the ‘ESTIMATE TEXT’ column.
The pivot puts the occupations as column heads, percentile values as as an index.
All selected occupations have values at 5 percentile levels (10, 25, 50, 75 & 90). Occupations with missing percentile values were excluded, a step that dropped the total number of occupations from about 93 to a plot-friendly value of 30.
The first plot looks at 3 occupations and the aggregated values for All workers (aggregation is part of the data set, was not done with python). At all percentile levels, there are more years of experience among Financial managers than among Bartenders. This plot used annotations to label each trace, while the actual legend is hidden.

image
image975×596 66.3 KB

The second plot includes all occupations that have values for all 5 percentiles. This screen shot is a bit noisy, however using plotly interactive legends you can select and de-select traces to find whatever

image
image975×437 119 KB

This dataset has so many dimensions to it. There are so many distinct and often unrelated attributes, with inconsistent units of measure. This challenged me well to come up with something useful to plot.

Here is the code

import plotly.express as px
import polars as pl
#-------------------------------------------------------------------------------
#    Read csv data to polars dataframe, clean the data.  
#-------------------------------------------------------------------------------
df_source = (
    pl.read_csv(
        'ors-limited-dataset.csv',
    )
    .filter(pl.col('CATEGORY') == 'Prior work experience')
    .filter(pl.col('ESTIMATE TEXT').str.contains('percentile'))
    # .filter(pl.col('OCCUPATION') != 'All workers')
    .with_columns(
        OCCUPATION = 
        pl.when(pl.col('OCCUPATION').str.starts_with('Executive secretaries'))
          .then(pl.lit('Exec Admin')).otherwise('OCCUPATION')
    )
    .with_columns(pl.col('ESTIMATE').cast(pl.Float32).cast(pl.Int16))
    .with_columns(MED_YEARS_EXP = pl.col('ESTIMATE')/365)
    .with_columns(PERCENTILE =  pl.lit(0))  # initialize all values to zero
    .with_columns(
        PERCENTILE = 
            pl.when(pl.col('ESTIMATE TEXT').str.contains('10'))
                .then(pl.lit(10))
                .when(pl.col('ESTIMATE TEXT').str.contains('25'))
                .then(pl.lit(25))
                .when(pl.col('ESTIMATE TEXT').str.contains('50'))
                .then(pl.lit(50))
                .when(pl.col('ESTIMATE TEXT').str.contains('75'))
                .then(pl.lit(75))
                .when(pl.col('ESTIMATE TEXT').str.contains('90'))
                .then(pl.lit(90))                                                
    )
    .pivot(
        on = 'OCCUPATION',
        index='PERCENTILE',
        values= 'MED_YEARS_EXP'
    )
    .sort('PERCENTILE')
)

#-------------------------------------------------------------------------------
#    Find columns that have data for all 5 percentiles, use for scatter plot
#-------------------------------------------------------------------------------
cols_w_all_data = sorted(
    [
        c for c in df_source.columns 
        if (len(df_source[c].drop_nulls()) == 5) 
    ]
)
fig=px.scatter(
    df_source,
    'PERCENTILE',
    cols_w_all_data,
    template='simple_white',
    title='Experience by Occupation'.upper(),
    width=1200, height=500
)
fig.update_layout(
    xaxis_title='Percentile'.upper(),
    yaxis_title='Required Years of Experience'.upper(),
    xaxis_showgrid=True,
    legend={'title':'Occupation'}
    )
fig.update_traces(mode='markers+lines')
fig.update_xaxes(
    tickmode = 'array',
    tickvals = [10, 25, 50, 75, 90],
    ticktext= [10, 25, 50, 75, 90],
    )

fig.show()

#-------------------------------------------------------------------------------
#    Make a clean plot showing 3 selected occupations, and All workrs
#-------------------------------------------------------------------------------
color_dict ={
    'All workers': px.colors.qualitative.G10[0], # px.colors.qualitative'red',
    'Bartenders':px.colors.qualitative.G10[1], #'green',
    'Exec Admin':px.colors.qualitative.G10[2], #'blue',
    'Financial managers':px.colors.qualitative.G10[3], #'yellow
}
plot_cols = sorted(color_dict.keys())
plot_colors = [color_dict.get(c) for c in plot_cols]

fig=px.scatter(
    df_source,
    'PERCENTILE',
    plot_cols,
    template='simple_white',
    title='Experience by Occupation'.upper(),
    width=800, height=500
)
fig.update_layout(
    xaxis_title='Percentile'.upper(),
    yaxis_title='Required Years of Experience'.upper(),
    xaxis_showgrid=True,
    showlegend=False
    )
fig.update_traces(mode='markers+lines')
fig.update_xaxes(
    tickmode = 'array',
    tickvals = [10, 25, 50, 75, 90],
    ticktext= [10, 25, 50, 75, 90],
    )
#-------------------------------------------------------------------------------
#    Touch up and annotate
#-------------------------------------------------------------------------------
annotation_x_offset = 1
for i in range(4):
    fig['data'][i]['line']['color']=plot_colors[i]
    fig['data'][i]['marker']['color']=plot_colors[i]
    x_annotate = annotation_x_offset + df_source['PERCENTILE'].to_list()[-1]
    y_annotate = df_source[plot_cols[i]].to_list()[-1]
    fig.add_annotation(
        x=x_annotate, xanchor='left',
        y=y_annotate, 
        showarrow=False,
        text=f'{plot_cols[i]}',
        align='right',
        font=dict(size=14, color=plot_colors[i]),
    )
fig.show()
5 Likes