![](\"https:\/\/storage.ning.com\/topology\/rest\/1.0\/file\/get\/3636885413?profile=RESIZE_710x\")
<\/a><\/h1>\n\u00a0<\/p>\n
\u00a0<\/p>\n
Introduction<\/h2>\n
\u00a0<\/p>\n
Co-relation does not equal causation<\/em><\/strong> \u2013 is a mantra drilled into a Data Scientist from an early age<\/p>\nThat\u2019s fine ..<\/p>\n
But very few talk of the follow-on question ..<\/p>\n
How exactly do you determine causation?<\/em><\/strong><\/p>\nThis problem is further compounded because most books and examples are based on standard datasets (ex: Boston, Iris etc) .<\/p>\n
These examples do not discuss causation because the features chosen are already determined to be causal (ex the factors affecting house prices are chosen to be causal)<\/p>\n
So, if we start from the beginning (without simplified examples) how do you know if a particular variable is a causal variable?<\/p>\n
Firstly, causality cannot be determined from data alone.<\/p>\n
Data gives co-relation, but data alone cannot determine causation<\/p>\n
To determine causation, we need to perform an experiment or a controlled study<\/strong><\/p>\nBackground<\/h2>\n
In a statistical sense, two or more variables are related if their values change correspondingly i.e. increase or decrease together. On the other hand, if there is a causal relationship between two variables, then the occurrence of one depends on the other i.e. they exhibit a cause and effect relationship. For example, smoking causes lung cancer is a causal relationship while smoking is correlated to alcoholism but does not cause alcoholism. \u00a0<\/p>\n
Correlation is typically measured using Pearson\u2019s coefficient or Spearman\u2019s coefficient. If there is correlation, then further investigation is needed to establish if there is a causal relationship.<\/p>\n
How can causation be established?<\/h2>\n
The most effective way of establishing causation is by means of a controlled study.<\/p>\n
In a controlled study, the sample or population is split in two, with both groups being comparable in almost every way.<\/p>\n
The two groups then receive different treatments, and the outcomes of each group are assessed.\u00a0<\/p>\n
For example, in medical research, one group is given a placebo whereas the other group is given a new medication.<\/p>\n
So, in a nutshell – “To find out what happens when you change something, it is necessary to change it.”…<\/em>There are things you learn from perturbing a system that you’ll never find out from any amount of passive observation.<\/p>\nSource: http:\/\/people.umass.edu\/~stanek\/pdffiles\/causal-holland.pdf<\/a><\/p>\n\u00a0<\/p>\n
The design of controlled experiments is a non-trivial exercise:<\/p>\n
\n- You may have measurement error problems<\/li>\n
- subjects might drop the study or not follow instructions, among other issues.<\/li>\n
- You will need to make assumptions about how things are related to determine inference.<\/li>\n
- You may have incomplete\/imprecise data<\/li>\n
- Target causal quantity of interest may not be well defined<\/li>\n
- Confounding variables. A confounder is a variable that influences both the dependent variable and independent variable, causing a spurious association.<\/li>\n
- Selection bias (self-selection, truncated samples)<\/li>\n
- Measurement error (that can induce confounding, not only noise)<\/li>\n
- Misspecification (e.g., wrong functional form)<\/li>\n
- External validity problems (wrong inference to target population)<\/li>\n<\/ul>\n
Adapted from source<\/a><\/p>\nFinally, there are some methods like the Granger causality<\/a> that is a statistical method which demonstrates some causality (with limitations)<\/p>\n\u00a0<\/p>\n
Sources<\/h2>\n
https:\/\/abs.gov.au\/websitedbs\/a3121120.nsf\/home\/statistical+language+-+correlation+and+causation<\/a><\/p>\nWhy do we need causality in data science<\/a><\/p>\nImage source: Khan academy<\/a><\/p>\n<\/div>\nGo to Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"
That\u2019s fine ..<\/p>\n
But very few talk of the follow-on question ..<\/p>\n
How exactly do you determine causation?<\/em><\/strong><\/p>\n This problem is further compounded because most books and examples are based on standard datasets (ex: Boston, Iris etc) .<\/p>\n These examples do not discuss causation because the features chosen are already determined to be causal (ex the factors affecting house prices are chosen to be causal)<\/p>\n So, if we start from the beginning (without simplified examples) how do you know if a particular variable is a causal variable?<\/p>\n Firstly, causality cannot be determined from data alone.<\/p>\n Data gives co-relation, but data alone cannot determine causation<\/p>\n To determine causation, we need to perform an experiment or a controlled study<\/strong><\/p>\n In a statistical sense, two or more variables are related if their values change correspondingly i.e. increase or decrease together. On the other hand, if there is a causal relationship between two variables, then the occurrence of one depends on the other i.e. they exhibit a cause and effect relationship. For example, smoking causes lung cancer is a causal relationship while smoking is correlated to alcoholism but does not cause alcoholism. \u00a0<\/p>\n Correlation is typically measured using Pearson\u2019s coefficient or Spearman\u2019s coefficient. If there is correlation, then further investigation is needed to establish if there is a causal relationship.<\/p>\n The most effective way of establishing causation is by means of a controlled study.<\/p>\n In a controlled study, the sample or population is split in two, with both groups being comparable in almost every way.<\/p>\n The two groups then receive different treatments, and the outcomes of each group are assessed.\u00a0<\/p>\n For example, in medical research, one group is given a placebo whereas the other group is given a new medication.<\/p>\n So, in a nutshell – “To find out what happens when you change something, it is necessary to change it.”…<\/em>There are things you learn from perturbing a system that you’ll never find out from any amount of passive observation.<\/p>\n Source: http:\/\/people.umass.edu\/~stanek\/pdffiles\/causal-holland.pdf<\/a><\/p>\n \u00a0<\/p>\n The design of controlled experiments is a non-trivial exercise:<\/p>\n Adapted from source<\/a><\/p>\n Finally, there are some methods like the Granger causality<\/a> that is a statistical method which demonstrates some causality (with limitations)<\/p>\n \u00a0<\/p>\n https:\/\/abs.gov.au\/websitedbs\/a3121120.nsf\/home\/statistical+language+-+correlation+and+causation<\/a><\/p>\n Why do we need causality in data science<\/a><\/p>\n Image source: Khan academy<\/a><\/p>\n<\/div>\n Go to Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"Background<\/h2>\n
How can causation be established?<\/h2>\n
\n
Sources<\/h2>\n