Data visualisation Challenges and opportunities in children's social care James Geddes | 10 December 2018 � 1
The start of a journey Local authority Note: values suppressed by LAs for confidentiality purposes were replaced with 3. The model used to produce the grey shade is a binomial distribution with rate set to the average returning � 3 home rate in the data.
If visualisation is the answer, what is the question? Image: Bruno Muria, deviantart � 4
Where should we start? Describe � 5
Where should we start? You've got to be very careful if you don't know where you are going, because you might not get there. Describe Yogi Berra (d. 2015) � 6
Where should we start? The philosophers have Act only interpreted the world, in various ways. The point, however, is to change it. Describe Karl Marx (d. 1883) � 7
� 8
Children's services Analysis Tool (ChAT) Page 4 Contacts in the last 3 months from 30/03/2017 Early Help in the last 6 months from 30/12/2016 to 30/06/2017 to 30/06/2017 185 contacts 204 Early Help Assessments, Targeted Interventions, and CAFs Contacts that also appear Early Help cases that also Age and gender Age and gender on the Referrals list appear on the Referrals list 75 Males (41%) 6 Unknown (3%) 109 Males (53%) 1 Unknown (0%) 104 Females (56%) Proportional 0-17 population 94 Females (46%) Proportional 0-17 population Yes No Yes No Aged 20+ Aged 20+ 15 15 10 10 47% 11% 5 5 Age 0 Age 0 15 10 5 0 5 10 15 10 5 0 5 10 15 10 5 0 5 10 15 10 5 0 5 10 Contact source Organisation completing assessment Contact source Referral comparison Individual Individual 5A - LA Services (Internal) 58% 12% 8% Schools Schools 28% 26% 6 - Police 20% Education services 0.0% Education services 0.0% 3D - Health Services (Primary) 7.8% Health services Health services 20% 0% 2A - Schools 7% Housing Housing 10.9% 4.9% 3A - Health Services (GP) 3.9% LA services 7% LA services 17% Police 3E - Health Services (A&E) 2% Police 15% 14% Other legal agency Other legal agency 4% 8% 0% Other 7% Other 6% 0% Anonymous 4.3% Anonymous 7.8% 0% Unknown Unknown 0.0% 0.0% Other 0.0% Not recorded Not recorded 0.0% 0.0% Children with more than 1 contact in period Ethnic backgrounds Children with more than 1 EH record in period Ethnic backgrounds 19 White 85% White 82% 1 6% 5% Mixed Mixed Asian or Asian British 1% Asian or Asian British 5% 6 Black or black British 0% Black or black British 7% Other ethnic group 1% Other ethnic group 0% 0 0 0 Not stated 7% Not stated 0% 0% 0% Not recorded Not recorded 2 contacts 3 contacts 4 or more 2 EHA/CAFs 3 EHA/CAFs 4 or more See page 20 for comparisons See page 20 for comparisons � 9
To make a decision we must know the outcome Act Predict Describe � 10
Example of "prediction" chart. Redacted. � 11 The Alan Turing Institute
To make a decision we must know the outcome Act Prediction is very difficult, especially about the future. Predict Describe Niels Bohr (d. 1962) (attr.) � 12
Note: chart shows the percentage of children who left care that were returned to parents or relatives for years 2011-2017 inclusive. Values suppressed for confidentiality purposes were replaced � 13 by 3. The model used to produce the grey shaded regions is a binomial likelihood function combined with a gaussian process, and is not final. Local authorities have been anonymised.
To predict we must have a model—in other words, a theory Act All models are wrong— but some are useful. Predict Explain Describe George Box (d. 2013) � 14
Image by DkEgy - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=61466910 � 15
Example of "small multiples" plot. Redacted. � 16
The steps of data science Act Focussed ? Predict Explain ? Exploratory Describe � 17
This chart raises questions ... Note: values suppressed by LAs for confidentiality purposes were replaced with 3. The model used to produce the grey shade is a binomial distribution with rate set to the average returning � 18 home rate in the data.
... which we hoped this chart might answer Local authority Note: Chart shows t-SNE dimensionality reduction of LAs based on the 22 demographic values and associated weights provided in the “Statistical neighbours benchmarking tool”. Plot made using scikit-learn with a particular choice of parameters. � 19
We ended up at the beginning It is a capital mistake to Act theorise before one has data. Predict Explain Describe � 20
Placeholder for Gapminder � 21
Image: shipmap.org � 22
Beginning again Key Front Door Team District Teams EH CIN Early Help Other Universal Support NFA/ cancel r e h m t O o r f C P C A L A L O C&F Ax Universal Support Initial Referral Early Help Contact Other Universal Targeted NFA/ EH IAG Strategy Section 47 ICPC Discussion NFA/ Other Other CP NFA/ CIN NFA/ CIN cancel cancel cancel Closure � 23
Thank you � 24
Recommend
More recommend
