4 ms·

Yes, nice catch! Thanks :) In case you're interested, we want to explore opportunities to expand the "grammar" of assumptions people might want to express. And

by emjun 7y ago

Yes, nice catch! Thanks :) In case you're interested, we want to explore opportunities to expand the "grammar" of assumptions people might want to express. And, that's our hope. We'd love to increase people's awareness and confidence in their analyses.

bitminer 7y ago
There are a few errors in the paper. "HCI" is not defined. Figure 2 refers to independent and dependant variables for an observational study but the text under "Study Design" says it should be "contributor" or "outcome". The software has zero releases. This doesn't allow comments w.r.t. versions and possible bug fixes. I like the concept. Adding a graphic representing the model would help make it interpretable by non-numerate stakeholders, such as management. Or politicians (!) It would be helpful to be able to make some statement on future support....is this going to be around for a few years?
- emjun 7y ago
  Thanks so much for pointing these out :) Yes, as we speak, I am working on updating the README, supporting docs, and a better website for the language/project. I will let you know once I have these up and running! It would be great to get your feedback. My plan is to continue building on Tea for at least a few more years and provide support for new use cases, statistical methods, etc. Are there topics/concerns you'd like to see addressed? :)
  - kyllo 7y ago
    I read and shared your paper and GitHub link with the rest of my data science team here at Microsoft Azure--we do a lot of hypothesis testing of this nature in our day-to-day work (using R or Python) and a tool like Tea would go a long way in helping us apply stats tests more efficiently and thoroughly. Good to know that you're planning to continue development, we'll be following with interest! Some other areas to potentially explore could be support for power analysis, sequential analysis and stopping rules, for use in the pre-trial phase of controlled experiments such as A/B tests. I'd also suggest adding a LICENSE file to your project to clarify how it's allowed to be used.

by emjun 7y ago

bitminer 7y ago

There are a few errors in the paper. "HCI" is not defined. Figure 2 refers to independent and dependant variables for an observational study but the text under "Study Design" says it should be "contributor" or "outcome". The software has zero releases. This doesn't allow comments w.r.t. versions and possible bug fixes. I like the concept. Adding a graphic representing the model would help make it interpretable by non-numerate stakeholders, such as management. Or politicians (!) It would be helpful to be able to make some statement on future support....is this going to be around for a few years?

emjun 7y ago
Thanks so much for pointing these out :) Yes, as we speak, I am working on updating the README, supporting docs, and a better website for the language/project. I will let you know once I have these up and running! It would be great to get your feedback. My plan is to continue building on Tea for at least a few more years and provide support for new use cases, statistical methods, etc. Are there topics/concerns you'd like to see addressed? :)
- kyllo 7y ago
  I read and shared your paper and GitHub link with the rest of my data science team here at Microsoft Azure--we do a lot of hypothesis testing of this nature in our day-to-day work (using R or Python) and a tool like Tea would go a long way in helping us apply stats tests more efficiently and thoroughly. Good to know that you're planning to continue development, we'll be following with interest! Some other areas to potentially explore could be support for power analysis, sequential analysis and stopping rules, for use in the pre-trial phase of controlled experiments such as A/B tests. I'd also suggest adding a LICENSE file to your project to clarify how it's allowed to be used.