Chapter 6 — College Scorecard (Python supplement)¶
Condensed notebook for the Python portion of Chapter 6: College Data.
Target graphics:
- SAT math vs. verbal scatter with trendline and fit summary (Figures 6.37–6.41)
- Cost vs. acceptance rate: scatter, Indiana highlight, labels, and graph-objects version (Figures 6.43–6.46)
Data: Download institution-level data from the College Scorecard and save Most-Recent-Cohorts-Institution_2022_23.csv in Data For Condensed Notebooks (or update the path in the first read_csv cell). Figures 6.1–6.3 in the book show the download site and dictionary.
Dependencies:
pandasplotlynumpystatsmodels(fortrendline='ols'in Plotly Express)
Imports and Plotly output¶
The chapter uses a PDF-capable default renderer (requires Kaleido if your environment uses that renderer).
In [1]:
import pandas as pd
import plotly.express as px
import numpy as np
In [2]:
# Set output options.
import plotly.io as pio
pio.renderers.default = "pdf+jupyterlab+notebook"
Reading and cleaning the data¶
Load the full Scorecard CSV, then restrict to the columns used in the chapter.
In [3]:
csc_all = pd.read_csv('../Data For Condensed Notebooks/Most-Recent-Cohorts-Institution_2022_23.csv')
/var/folders/_x/nbw2t83x75d0lr3jn4pg_cx00000gn/T/ipykernel_26711/1372049275.py:1: DtypeWarning: Columns (0: NPCURL, 1: INC_N, 2: DEP_STAT_N, 3: APPL_SCH_N, 4: AGE_ENTRY, 5: AGEGE24, 6: FAMINC, 7: MD_FAMINC, 8: PCT_WHITE, 9: PCT_BLACK, 10: PCT_ASIAN, 11: PCT_HISPANIC, 12: PCT_BA, 13: PCT_GRAD_PROF, 14: PCT_BORN_US, 15: MEDIAN_HH_INC, 16: POVERTY_RATE, 17: UNEMP_RATE, 18: LN_MEDIAN_HH_INC, 19: MN_EARN_WNE_P7, 20: GT_25K_P7, 21: PCT10_EARN_WNE_P8, 22: PCT90_EARN_WNE_P8, 23: C150_L4_POOLED_SUPP, 24: C150_4_POOLED_SUPP, 25: C200_L4_POOLED_SUPP, 26: C200_4_POOLED_SUPP, 27: ALIAS, 28: T4APPROVALDATE, 29: RET_FT4_POOLED_SUPP, 30: RET_FTL4_POOLED_SUPP, 31: RET_PT4_POOLED_SUPP, 32: RET_PTL4_POOLED_SUPP, 33: TRANS_4_POOLED_SUPP, 34: TRANS_L4_POOLED_SUPP, 35: C100_4_POOLED_SUPP, 36: C100_L4_POOLED_SUPP, 37: OMAWDP6_FTFT_POOLED_SUPP, 38: OMAWDP8_FTFT_POOLED_SUPP, 39: OMENRYP8_FTFT_POOLED_SUPP, 40: OMENRAP8_FTFT_POOLED_SUPP, 41: OMENRUP8_FTFT_POOLED_SUPP, 42: OMAWDP6_PTFT_POOLED_SUPP, 43: OMAWDP8_PTFT_POOLED_SUPP, 44: OMENRYP8_PTFT_POOLED_SUPP, 45: OMENRAP8_PTFT_POOLED_SUPP, 46: OMENRUP8_PTFT_POOLED_SUPP, 47: OMAWDP6_FTNFT_POOLED_SUPP, 48: OMAWDP8_FTNFT_POOLED_SUPP, 49: OMENRYP8_FTNFT_POOLED_SUPP, 50: OMENRAP8_FTNFT_POOLED_SUPP, 51: OMENRUP8_FTNFT_POOLED_SUPP, 52: OMAWDP6_PTNFT_POOLED_SUPP, 53: OMAWDP8_PTNFT_POOLED_SUPP, 54: OMENRYP8_PTNFT_POOLED_SUPP, 55: OMENRAP8_PTNFT_POOLED_SUPP, 56: OMENRUP8_PTNFT_POOLED_SUPP, 57: CIPTITLE2, 58: CIPTITLE3, 59: CIPTITLE4, 60: CIPTITLE5, 61: CIPTITLE6, 62: OMENRYP_ALL_POOLED_SUPP, 63: OMENRAP_ALL_POOLED_SUPP, 64: OMAWDP8_ALL_POOLED_SUPP, 65: OMENRUP_ALL_POOLED_SUPP, 66: OMENRYP_FIRSTTIME_POOLED_SUPP, 67: OMENRAP_FIRSTTIME_POOLED_SUPP, 68: OMAWDP8_FIRSTTIME_POOLED_SUPP, 69: OMENRUP_FIRSTTIME_POOLED_SUPP, 70: OMENRYP_NOTFIRSTTIME_POOLED_SUPP, 71: OMENRAP_NOTFIRSTTIME_POOLED_SUPP, 72: OMAWDP8_NOTFIRSTTIME_POOLED_SUPP, 73: OMENRUP_NOTFIRSTTIME_POOLED_SUPP, 74: OMENRYP_FULLTIME_POOLED_SUPP, 75: OMENRAP_FULLTIME_POOLED_SUPP, 76: OMAWDP8_FULLTIME_POOLED_SUPP, 77: OMENRUP_FULLTIME_POOLED_SUPP, 78: OMENRYP_PARTTIME_POOLED_SUPP, 79: OMENRAP_PARTTIME_POOLED_SUPP, 80: OMAWDP8_PARTTIME_POOLED_SUPP, 81: OMENRUP_PARTTIME_POOLED_SUPP, 82: FTFTPCTPELL_POOLED_SUPP, 83: FTFTPCTFLOAN_POOLED_SUPP, 84: LPSTAFFORD_CNT, 85: LPSTAFFORD_AMT, 86: C150_L4_PELL_POOLED_SUPP, 87: C150_4_PELL_POOLED_SUPP, 88: OMAWDP8_PELL_FTFT_POOLED_SUPP, 89: OMENRYP8_PELL_FTFT_POOLED_SUPP, 90: OMENRAP8_PELL_FTFT_POOLED_SUPP, 91: OMENRUP8_PELL_FTFT_POOLED_SUPP, 92: OMAWDP8_PELL_PTFT_POOLED_SUPP, 93: OMENRYP8_PELL_PTFT_POOLED_SUPP, 94: OMENRAP8_PELL_PTFT_POOLED_SUPP, 95: OMENRUP8_PELL_PTFT_POOLED_SUPP, 96: OMAWDP8_PELL_FTNFT_POOLED_SUPP, 97: OMENRYP8_PELL_FTNFT_POOLED_SUPP, 98: OMENRAP8_PELL_FTNFT_POOLED_SUPP, 99: OMENRUP8_PELL_FTNFT_POOLED_SUPP, 100: OMAWDP8_PELL_PTNFT_POOLED_SUPP, 101: OMENRYP8_PELL_PTNFT_POOLED_SUPP, 102: OMENRAP8_PELL_PTNFT_POOLED_SUPP, 103: OMENRUP8_PELL_PTNFT_POOLED_SUPP, 104: OMENRYP_PELL_ALL_POOLED_SUPP, 105: OMENRAP_PELL_ALL_POOLED_SUPP, 106: OMAWDP8_PELL_ALL_POOLED_SUPP, 107: OMENRUP_PELL_ALL_POOLED_SUPP, 108: OMENRYP_PELL_FTT_POOLED_SUPP, 109: OMENRAP_PELL_FTT_POOLED_SUPP, 110: OMAWDP8_PELL_FTT_POOLED_SUPP, 111: OMENRUP_PELL_FTT_POOLED_SUPP, 112: OMENRYP_PELL_NFT_POOLED_SUPP, 113: OMENRAP_PELL_NFT_POOLED_SUPP, 114: OMAWDP8_PELL_NFT_POOLED_SUPP, 115: OMENRUP_PELL_NFT_POOLED_SUPP, 116: OMENRYP_PELL_FT_POOLED_SUPP, 117: OMENRAP_PELL_FT_POOLED_SUPP, 118: OMAWDP8_PELL_FT_POOLED_SUPP, 119: OMENRUP_PELL_FT_POOLED_SUPP, 120: OMENRYP_PELL_PT_POOLED_SUPP, 121: OMENRAP_PELL_PT_POOLED_SUPP, 122: OMAWDP8_PELL_PT_POOLED_SUPP, 123: OMENRUP_PELL_PT_POOLED_SUPP, 124: GT_THRESHOLD_P6_SUPP, 125: ADM_RATE_SUPP, 126: ADDR, 127: PCTPELL_DCS_POOLED_SUPP, 128: PCTFLOAN_DCS_POOLED_SUPP) have mixed types. Specify dtype option on import or set low_memory=False.
csc_all = pd.read_csv('../Data For Condensed Notebooks/Most-Recent-Cohorts-Institution_2022_23.csv')
You may see a DtypeWarning about mixed types: many columns contain strings such as NA or PS, so pandas cannot infer numeric dtypes for the whole file. The analyses below still work on the columns we keep; narrowing columns first is the usual fix. For Tableau, exporting clean columns (below) helps.
Check memory use of the full table.
In [4]:
csc_all.info()
<class 'pandas.DataFrame'> RangeIndex: 6484 entries, 0 to 6483 Columns: 3305 entries, UNITID to MD_EARN_WNE_MALE1_P11 dtypes: float64(919), int64(14), object(79), str(2293) memory usage: 163.5+ MB
Subset to OPEID, institution name, SAT 25th percentiles, admission rate, sticker price, city, and state.
In [5]:
csc = csc_all[['OPEID', 'INSTNM', 'SATMT25', 'SATVR25', 'ADM_RATE_ALL', 'COSTT4_A', 'CITY', 'STABBR']].copy()
Note: You could instead pass usecols=[...] to read_csv to read only these columns.
Inspect dtypes for the subset.
In [6]:
csc.dtypes
Out[6]:
OPEID float64 INSTNM str SATMT25 float64 SATVR25 float64 ADM_RATE_ALL float64 COSTT4_A float64 CITY str STABBR str dtype: object
Numeric columns are floats despite NA/PS strings elsewhere in the file. Inspect unique values for one SAT column.
In [7]:
csc['SATMT25'].unique()
Out[7]:
array([400., 590., nan, 613., 399., 560., 500., 610., 480., 460., 450.,
530., 493., 475., 570., 490., 510., 380., 423., 458., 553., 470.,
545., 445., 540., 600., 730., 568., 760., 650., 680., 635., 750.,
550., 740., 580., 640., 620., 440., 660., 410., 630., 382., 710.,
430., 505., 420., 520., 700., 425., 525., 398., 670., 563., 495.,
390., 593., 448., 562., 555., 443., 469., 473., 455., 558., 483.,
438., 478., 585., 485., 690., 720., 488., 533., 780., 477., 528.,
790., 515., 518., 482., 532., 225., 537., 603., 465., 535., 770.,
310., 598., 360., 565., 733., 340., 543., 350., 538., 463., 408.,
435., 618., 523., 583., 503., 393., 715., 648., 513., 370., 428.,
368., 422.])
read_csv maps the literal string "NA" to NaN for these columns. Cast ID and text fields to pandas string dtype.
In [8]:
for c in ['OPEID', 'INSTNM', 'CITY', 'STABBR']:
csc[c] = csc[c].astype('string')
csc.dtypes
Out[8]:
OPEID string INSTNM string SATMT25 float64 SATVR25 float64 ADM_RATE_ALL float64 COSTT4_A float64 CITY string STABBR string dtype: object
Write a small CSV for Tableau (Figure 6.18 / clean data workflow in the chapter).
In [9]:
csc.to_csv('../Data For Condensed Notebooks/csc_small.csv', index=False)
SAT scores¶
Compare math and verbal SAT (25th percentile) across institutions. Figure 6.37 — quick pandas scatter; 6.38–6.41 — interactive Plotly, trendline, and regression summary.
Preview the cleaned dataframe.
In [10]:
csc
Out[10]:
| OPEID | INSTNM | SATMT25 | SATVR25 | ADM_RATE_ALL | COSTT4_A | CITY | STABBR | |
|---|---|---|---|---|---|---|---|---|
| 0 | 100200.0 | Alabama A & M University | 400.0 | 430.0 | 0.683956 | 23167.0 | Normal | AL |
| 1 | 105200.0 | University of Alabama at Birmingham | 590.0 | 610.0 | 0.866794 | 26257.0 | Birmingham | AL |
| 2 | 2503400.0 | Amridge University | NaN | NaN | NaN | NaN | Montgomery | AL |
| 3 | 105500.0 | University of Alabama in Huntsville | 613.0 | 613.0 | 0.781043 | 25777.0 | Huntsville | AL |
| 4 | 100500.0 | Alabama State University | 399.0 | 429.0 | 0.965978 | 21900.0 | Montgomery | AL |
| ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 6479 | 4270802.0 | Wilton Simpson Technical College | NaN | NaN | NaN | NaN | Brooksville | FL |
| 6480 | 2609404.0 | Valley College - Fairlawn - School of Nursing | NaN | NaN | NaN | NaN | Fairlawn | OH |
| 6481 | 4247201.0 | Western Maricopa Education Center - Southwest ... | NaN | NaN | NaN | NaN | Buckeye | AZ |
| 6482 | 4247202.0 | Western Maricopa Education Center - Northeast ... | NaN | NaN | NaN | NaN | Phoenix | AZ |
| 6483 | 4285601.0 | Burlington County Institute of Technology - Ad... | NaN | NaN | NaN | NaN | Medford | NJ |
6484 rows × 8 columns
Figure 6.37 — Built-in pandas scatter (non-interactive).
In [11]:
csc.plot.scatter('SATMT25', 'SATVR25')
Out[11]:
<Axes: xlabel='SATMT25', ylabel='SATVR25'>
Identify the low corner outlier with idxmin (cf. Figure 6.10 in the Excel walkthrough).
In [12]:
min_index = csc['SATMT25'].idxmin()
csc.loc[min_index]
Out[12]:
OPEID 243300.0 INSTNM Rust College SATMT25 225.0 SATVR25 215.0 ADM_RATE_ALL 0.787596 COSTT4_A 16700.0 CITY Holly Springs STABBR MS Name: 1685, dtype: object
Figure 6.38 — Interactive plotly.express scatter with institution names on hover.
In [13]:
sat_fig = px.scatter(csc,
x='SATMT25',
y='SATVR25',
hover_data='INSTNM')
sat_fig
Figure 6.39 — OLS trendline (trendline='ols'). Requires statsmodels (uv add statsmodels or conda install -c conda-forge statsmodels); restart kernel after install if needed.
In [14]:
sat_fig = px.scatter(csc,
x='SATMT25',
y='SATVR25',
hover_data='INSTNM',
trendline='ols')
sat_fig
Figure 6.40 — Trendline in a contrasting color.
In [15]:
sat_fig = px.scatter(csc,
x='SATMT25',
y='SATVR25',
hover_data='INSTNM',
trendline='ols',
trendline_color_override='black')
sat_fig
Figure 6.41 — Dashed trendline via the figure’s trace objects.
In [16]:
# Change the line style of the trendline.
sat_fig.data[1].line.dash = 'dash'
sat_fig
Fit summary from px.get_trendline_results (see Plotly docs). The chapter discusses $R^2 \approx 0.90$ for this relationship.
In [17]:
ols_info = px.get_trendline_results(sat_fig)
ols_info
Out[17]:
| px_fit_results | |
|---|---|
| 0 | <statsmodels.regression.linear_model.Regressio... |
Access the underlying statsmodels summary for the first (only) trendline.
In [18]:
ols_info.px_fit_results.iloc[0].summary()
Out[18]:
| Dep. Variable: | y | R-squared: | 0.903 |
|---|---|---|---|
| Model: | OLS | Adj. R-squared: | 0.902 |
| Method: | Least Squares | F-statistic: | 9322. |
| Date: | Thu, 03 Sep 2026 | Prob (F-statistic): | 0.00 |
| Time: | 15:20:56 | Log-Likelihood: | -4636.6 |
| No. Observations: | 1008 | AIC: | 9277. |
| Df Residuals: | 1006 | BIC: | 9287. |
| Df Model: | 1 | ||
| Covariance Type: | nonrobust |
| coef | std err | t | P>|t| | [0.025 | 0.975] | |
|---|---|---|---|---|---|---|
| const | 71.8545 | 4.883 | 14.714 | 0.000 | 62.272 | 81.437 |
| x1 | 0.8796 | 0.009 | 96.550 | 0.000 | 0.862 | 0.898 |
| Omnibus: | 108.921 | Durbin-Watson: | 1.955 |
|---|---|---|---|
| Prob(Omnibus): | 0.000 | Jarque-Bera (JB): | 704.797 |
| Skew: | -0.229 | Prob(JB): | 9.02e-154 |
| Kurtosis: | 7.071 | Cond. No. | 3.45e+03 |
Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 3.45e+03. This might indicate that there are
strong multicollinearity or other numerical problems.
Strong linear association between math and verbal 25th percentiles; most table output is standard regression diagnostics.
Cost vs. acceptance rate¶
Figures 6.43–6.46 — Exploratory summaries, scatter, Indiana highlight, labels, and a graph-objects version.
In [19]:
csc['COSTT4_A'].describe()
Out[19]:
count 3316.000000 mean 29298.406514 std 17702.594870 min 5500.000000 25% 15092.750000 50% 23940.500000 75% 38900.500000 max 86964.000000 Name: COSTT4_A, dtype: float64
In [20]:
csc['ADM_RATE_ALL'].describe()
Out[20]:
count 2215.000000 mean 0.715795 std 0.234499 min 0.000000 25% 0.596295 50% 0.769961 75% 0.888959 max 1.000000 Name: ADM_RATE_ALL, dtype: float64
Counts differ between cost and admission columns—summary stats describe different sets of non-missing schools.
Most expensive institution by sticker price (idxmax).
In [21]:
csc.loc[csc['COSTT4_A'].idxmax()][['INSTNM', 'COSTT4_A']]
Out[21]:
INSTNM Jewish Theological Seminary of America COSTT4_A 86964.0 Name: 2093, dtype: object
Search institution names containing “Ohio” to see naming conventions.
In [22]:
csc[csc['INSTNM'].str.contains('Ohio')][['INSTNM', 'CITY']]
Out[22]:
| INSTNM | CITY | |
|---|---|---|
| 2407 | Athenaeum of Ohio | Cincinnati |
| 2422 | Central Ohio Technical College | Newark |
| 2433 | Ohio Christian University | Circleville |
| 2482 | Ohio Business College-Sheffield | Sheffield Village |
| 2483 | Ohio Business College-Sandusky | Sandusky |
| 2490 | Mercy College of Ohio | Toledo |
| 2491 | Methodist Theological School in Ohio | Delaware |
| 2509 | Northeast Ohio Medical University | Rootstown |
| 2510 | University of Northwestern Ohio | Lima |
| 2512 | Ohio Technical College | Cleveland |
| 2513 | Ohio Dominican University | Columbus |
| 2514 | Ohio Northern University | Ada |
| 2515 | Ohio State University Agricultural Technical I... | Wooster |
| 2516 | Ohio State University-Lima Campus | Lima |
| 2517 | Ohio State University-Mansfield Campus | Mansfield |
| 2518 | Ohio State University-Marion Campus | Marion |
| 2519 | Ohio State University-Newark Campus | Newark |
| 2520 | Ohio State Beauty Academy | Lima |
| 2521 | Ohio State College of Barber Styling | Columbus |
| 2523 | Ohio State School of Cosmetology-Canal Winchester | CANAL WINCHESTER |
| 2524 | Ohio State University-Main Campus | Columbus |
| 2525 | Ohio University-Eastern Campus | Saint Clairsville |
| 2526 | Ohio University-Chillicothe Campus | Chillicothe |
| 2527 | Ohio University-Southern Campus | Ironton |
| 2528 | Ohio University-Lancaster Campus | Lancaster |
| 2529 | Ohio University-Main Campus | Athens |
| 2530 | Ohio University-Zanesville Campus | Zanesville |
| 2531 | East Ohio College | East Liverpool |
| 2533 | Ohio Wesleyan University | Delaware |
| 3965 | Ohio State School of Cosmetology-Heath | Heath |
| 4013 | Ohio Media School-Valley View | Valley View |
| 4083 | Ohio Media School-Cincinnati | Norwood |
| 4710 | Ohio Media School-Columbus | Columbus |
| 4711 | Ohio Medical Career College | Dayton |
| 4714 | Chamberlain University-Ohio | Columbus |
| 5286 | DeVry University-Ohio | Columbus |
| 5322 | Ohio Institute of Allied Health | Huber Heights |
| 5916 | Ohio Business College-Dayton-Driving Academy | Trotwood |
Query a specific campus (quote style matters inside query).
In [23]:
csc.query('INSTNM == "Ohio State University-Main Campus"')[['INSTNM', 'COSTT4_A', 'ADM_RATE_ALL']]
Out[23]:
| INSTNM | COSTT4_A | ADM_RATE_ALL | |
|---|---|---|---|
| 2524 | Ohio State University-Main Campus | 28133.0 | 0.527236 |
Figure 6.43 — Scatter of sticker price vs. acceptance rate.
In [24]:
adm_cost_fig = px.scatter(csc,
x='ADM_RATE_ALL',
y='COSTT4_A',
hover_data='INSTNM')
adm_cost_fig
Figure 6.44 — Boolean mask for Indiana (STABBR == 'IN') with color and symbol encoding.
In [25]:
# Create a boolean DataSeries to identify Indiana schools.
IN_bool = csc['STABBR'] == 'IN'
In [26]:
adm_cost_fig = px.scatter(csc,
x='ADM_RATE_ALL',
y='COSTT4_A',
hover_data='INSTNM',
color=IN_bool,
symbol_sequence=['x-thin', 'circle'],
symbol=IN_bool
)
adm_cost_fig
Figure 6.45 — Label only selected Indiana schools; widen the x-axis and position text so labels are readable.
In [27]:
# Make a list of the institutions we want to label.
schools_to_label = ['Valparaiso University',
'University of Notre Dame',
'Purdue University-Main Campus',
'Veritas Baptist College',
'DePauw University',
'Earlham College',
'Indiana University-Bloomington',
'University of Indianapolis']
Build a labels column: empty except for chosen names (isin).
In [28]:
# Get the indexes of the rows for the schools we
# want to label.
idx = csc['INSTNM'].isin(schools_to_label)
# Create a new column with empty labels.
csc['labels'] = ''
# Replace the blank labels with school names
# for the chosen schools.
csc.loc[idx, 'labels'] = csc.loc[idx, 'INSTNM']
# Check that the labels column has the correct values in it.
csc['labels'].value_counts()
Out[28]:
labels
6476
DePauw University 1
Earlham College 1
University of Indianapolis 1
Indiana University-Bloomington 1
University of Notre Dame 1
Valparaiso University 1
Purdue University-Main Campus 1
Veritas Baptist College 1
Name: count, dtype: int64
Express version with text labels and axis titles.
In [29]:
# Create the scatter plot.
adm_cost_fig = px.scatter(csc,
x='ADM_RATE_ALL',
y='COSTT4_A',
hover_data='INSTNM',
color=IN_bool,
symbol_sequence=['x-thin', 'circle'],
symbol=IN_bool,
text='labels',
labels={'ADM_RATE_ALL': 'Acceptance Rate',
'COSTT4_A': 'Cost to Attend'},
height=600
)
# Remove the legend.
adm_cost_fig.update_layout(showlegend=False)
# Make sure none of the labels get cut off.
adm_cost_fig.update_xaxes(range=[-0.1, 1.15])
# Set where the labels appear relative to the points.
adm_cost_fig.update_traces(textposition='top center')
adm_cost_fig
Graph objects approach¶
Figure 6.46 — Two traces (non-Indiana vs. Indiana), explicit colors, opacity, and legend. Trace order puts Indiana on top.
In [30]:
# Import the graph objects part of Plotly.
import plotly.graph_objects as go
import numpy as np
Split the dataframe with IN_bool and np.invert.
In [31]:
# Pick out just the schools in Indiana.
IN_csc = csc[IN_bool].copy()
# Create a DataFrame of all other schools (not in Indiana).
non_IN_csc = csc[np.invert(IN_bool)].copy()
Add the same labels column on the Indiana subset only.
In [32]:
# Get the indexes of the rows for the schools we
# want to label.
idx = IN_csc['INSTNM'].isin(schools_to_label)
# Create a new column with empty labels.
IN_csc['labels'] = ''
# Replace the blank labels with school names
# for the chosen schools.
IN_csc.loc[idx, 'labels'] = IN_csc.loc[idx, 'INSTNM']
# Check that the labels column has the correct values in it.
IN_csc['labels'].value_counts()
Out[32]:
labels
134
DePauw University 1
Earlham College 1
University of Indianapolis 1
Indiana University-Bloomington 1
University of Notre Dame 1
Valparaiso University 1
Purdue University-Main Campus 1
Veritas Baptist College 1
Name: count, dtype: int64
go.Scatter: red markers + text for Indiana, translucent green × for others. If a label fails to appear, texttemplate='%{text}' can help (see chapter note).
In [33]:
# Make red dots for the Indiana schools.
# Add text labels for selected schools.
IN_trace = go.Scatter(x=IN_csc['ADM_RATE_ALL'],
y=IN_csc['COSTT4_A'],
mode='markers+text',
marker_color='red',
text=IN_csc['labels'],
textposition='top center',
texttemplate='%{text}',
name='Schools in Indiana',
hovertext=IN_csc['INSTNM']
)
# Make translucent green x's for non-Indiana schools.
non_IN_trace = go.Scatter(x=non_IN_csc['ADM_RATE_ALL'],
y=non_IN_csc['COSTT4_A'],
mode='markers',
marker_symbol='x',
marker_color='green',
marker_opacity=0.2,
name='Schools Outside Indiana',
hovertext=non_IN_csc['INSTNM']
)
# Create a figure with those traces, making sure IN schools
# are on top.
adm_cost_fig = go.Figure(data=[non_IN_trace, IN_trace])
# Add some formatting.
adm_cost_fig.update_layout(height=600, width=900)
adm_cost_fig.update_xaxes(range=[-0.1, 1.15], title='Acceptance Rate')
adm_cost_fig.update_yaxes(range=[8000, 85000], title='Cost to Attend')
adm_cost_fig.update_traces(textposition='top center')
adm_cost_fig
Graph objects give direct control over styling and the legend compared to the express shortcut above.
Caveat: “Cost to attend” here is sticker price; acceptance rates are hard to compare across institutions and over time. See the closing discussion in the chapter.