Tea Survey

%load_ext autoreload
%autoreload 2
%matplotlib inline

The autoreload extension is already loaded. To reload it, use:
  %reload_ext autoreload

Imports

import warnings
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from fastdownload import download_url
from IPython.display import display, Image
from aiking.data.external import * #We need to import this after fastai modules

from scipy import stats
sns.set(color_codes=True)
sns.set_palette(sns.color_palette("muted"))
warnings.filterwarnings("ignore")

Dataset Review

list_ds()

(#3) ['oxford-iiit-pet','california-housing-prices','mktr']

ds = get_ds('mktr')
(ds/"conjoint_analysis"/"tea").ls()

(#6) [Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/filled_conjoint_survey_tea.csv'),Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/partwort_utilities.csv'),Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/importance_ratings.csv'),Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/segmentation_level_partworths.csv'),Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/segmentation_importance_ratings.csv'),Path('/media/sf_gdrive/PPV/S_Personal_Study/aiking/data/mktr/conjoint_analysis/tea/respondent_demographic_sample.csv')]

path = (ds/"conjoint_analysis"/"tea")
data = pd.read_csv(path/"filled_conjoint_survey_tea.csv")
data

	profile	price_per_kg	variety	kind	aroma	res1	res2	res3	res4	res5	...	res91	res92	res93	res94	res95	res96	res97	res98	res99	res100
0	profile1	250	black	bags	yes	8	0	4	6	5	...	3	9	5	9	10	8	5	5	7	9
1	profile2	500	green	bags	yes	1	10	10	7	1	...	7	3	4	3	8	9	3	3	4	7
2	profile3	250	red	bags	yes	1	3	3	4	7	...	5	3	8	3	0	2	0	1	3	4
3	profile4	750	black	granulated	yes	3	5	5	9	8	...	5	5	1	5	0	6	9	4	9	10
4	profile5	500	red	granulated	yes	9	1	4	6	6	...	3	1	2	0	2	6	0	3	0	9
5	profile6	500	black	leafy	yes	2	4	1	3	10	...	5	2	2	4	0	2	8	8	5	3
6	profile7	750	green	leafy	yes	7	8	2	7	7	...	3	0	4	0	0	1	0	5	4	2
7	profile8	250	red	leafy	yes	2	6	0	4	10	...	9	0	3	0	0	2	0	9	0	1
8	profile9	500	black	bags	no	2	2	0	8	6	...	3	1	3	4	0	2	6	6	5	2
9	profile10	750	red	bags	no	2	9	1	5	6	...	9	0	1	0	0	2	0	8	0	3
10	profile11	250	green	granulated	no	2	7	8	2	6	...	5	0	2	6	0	7	9	4	6	9
11	profile12	250	black	leafy	no	3	5	9	10	10	...	4	2	8	2	6	6	2	3	10	10
12	profile13	500	red	leafy	no	4	2	7	9	7	...	4	1	6	0	3	6	0	1	8	8

13 rows × 105 columns

pdt_attributes = ['price_per_kg', 'variety', 'kind', 'aroma']
profiles = ['profile']

resps = data.columns.difference(list(pdt_attributes+profiles)); resps

Index(['res1', 'res10', 'res100', 'res11', 'res12', 'res13', 'res14', 'res15',
       'res16', 'res17', 'res18', 'res19', 'res2', 'res20', 'res21', 'res22',
       'res23', 'res24', 'res25', 'res26', 'res27', 'res28', 'res29', 'res3',
       'res30', 'res31', 'res32', 'res33', 'res34', 'res35', 'res36', 'res37',
       'res38', 'res39', 'res4', 'res40', 'res41', 'res42', 'res43', 'res44',
       'res45', 'res46', 'res47', 'res48', 'res49', 'res5', 'res50', 'res51',
       'res52', 'res53', 'res54', 'res55', 'res56', 'res57', 'res58', 'res59',
       'res6', 'res60', 'res61', 'res62', 'res63', 'res64', 'res65', 'res66',
       'res67', 'res68', 'res69', 'res7', 'res70', 'res71', 'res72', 'res73',
       'res74', 'res75', 'res76', 'res77', 'res78', 'res79', 'res8', 'res80',
       'res81', 'res82', 'res83', 'res84', 'res85', 'res86', 'res87', 'res88',
       'res89', 'res9', 'res90', 'res91', 'res92', 'res93', 'res94', 'res95',
       'res96', 'res97', 'res98', 'res99'],
      dtype='object')

attr_levels = {i:data[i].unique() for i in pdt_attributes}; attr_levels

{'price_per_kg': array([250, 500, 750]),
 'variety': array(['black', 'green', 'red'], dtype=object),
 'kind': array(['bags', 'granulated', 'leafy'], dtype=object),
 'aroma': array(['yes', 'no'], dtype=object)}

data[profiles+list(resps)].set_index('profile').T.describe().T

	count	mean	std	min	25%	50%	75%	max
profile
profile1	100.0	5.99	3.270607	0.0	4.00	7.0	8.25	10.0
profile2	100.0	4.17	3.455504	0.0	1.00	3.0	7.00	10.0
profile3	100.0	2.68	2.273608	0.0	1.00	2.5	4.00	10.0
profile4	100.0	4.73	3.011275	0.0	2.75	5.0	6.50	10.0
profile5	100.0	3.55	3.182576	0.0	1.00	2.0	6.00	9.0
profile6	100.0	3.36	2.642007	0.0	2.00	3.0	5.00	10.0
profile7	100.0	2.53	2.572013	0.0	0.00	2.0	4.00	8.0
profile8	100.0	2.40	2.759739	0.0	0.00	2.0	3.00	10.0
profile9	100.0	2.55	2.253505	0.0	1.00	2.0	4.00	8.0
profile10	100.0	2.35	2.668087	0.0	0.00	2.0	3.00	9.0
profile11	100.0	4.45	2.709411	0.0	2.00	4.0	6.00	10.0
profile12	100.0	5.01	3.625095	0.0	1.75	4.0	9.00	10.0
profile13	100.0	3.79	3.009396	0.0	1.00	3.5	6.00	9.0

Note

Above table describes average rating for 13 different profiles for the product by 100 respondents

ratings = np.array(data[resps])
ratings.min(), ratings.max(), ratings.mean(), ratings.std()

(0, 10, 3.6584615384615384, 3.101160691279331)

sns.histplot(np.ravel(ratings))

<AxesSubplot:ylabel='Count'>