Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

90% of the T Distribution

entropicthoughts.com

Post reply on HN
Print view View original

11–20 of 20 posts

Re: 90% of the T Distribution

#11

Post by kqr »

srean wrote: post #10
Fraterkes wrote: post #2

This is neither here nor there: I was reading the about page of the author, and it contains a passage that slightly confused me: "My name is Chris and I live in Sweden. I have a beautiful, supportive wife whose love I will never be able to requite, neither in degree nor kind." English isn't my first language, how should the second sentence be interpreted?

To address the degree bit, one could bake a couple of pies once in a while. Let me show myself out.

This is good on many levels. It hits temperature degrees, angular degrees, and it's actually accurate advice.

Re: 90% of the T Distribution

#12

Post by srean »

kqr wrote: post #11
srean wrote: post #10

Earlier quoted context omitted.

To address the degree bit, one could bake a couple of pies once in a while. Let me show myself out.

This is good on many levels. It hits temperature degrees, angular degrees, and it's actually accurate advice.

Was entirely tongue in cheek. A long time admirer of your blog.

Having a wonderful wife myself I can understand the feelings.

Re: 90% of the T Distribution

#13

Post by kqr »

Author here. It feels a bit pushy and uncharacteristic for me, but now that this ended up on the front page I would really appreciate if I could get 3.5 minutes out of your life, by taking this keyboard latency probe: https://xkqr.org/bellwether/keyboardtest.html

I've already received very good data, but there are some tenuous connections that are still too noisy to be certain of

- Are Apple keyboards really slower that non-Apple keyboards?

- Are cheap keyboards really faster than expensive keyboards?

- Are virtually all split keyboards programmable?

More samples would help nail these things down. I'll share the analysis with the community once I'm done, of course. (The analysis pipeline is mostly automated so I can work on analysis in parallel with receiving more submissions.)

Re: 90% of the T Distribution

#14

Post by ranma42 »

kqr wrote: post #13

Author here. It feels a bit pushy and uncharacteristic for me, but now that this ended up on the front page I would really appreciate if I could get 3.5 minutes out of your life, by taking this keyboard latency probe: https://xkqr.org/bellwether/keyboardtest.html I've already received very good data, but there are some tenuous connections that are still too noisy to be certain of - Are Apple keyboards really slower t…

How do you distinguish between input (Keyboard/USB) and output (Browser/Graphics stack) delays in this test?

Re: 90% of the T Distribution

#15

Post by kqr »

ranma42 wrote: post #14
kqr wrote: post #13

Author here. It feels a bit pushy and uncharacteristic for me, but now that this ended up on the front page I would really appreciate if I could get 3.5 minutes out of your life, by taking this keyboard latency probe: https://xkqr.org/bellwether/keyboardtest.html I've already received very good data, but there are some tenuous connections that are still too noisy to be certain of - Are Apple keyboards really slower t…

How do you distinguish between input (Keyboard/USB) and output (Browser/Graphics stack) delays in this test?

That's outside the scope of what I wanted to find out so my plan was to treat that as part of the random effects varying from user to user.

But now that you say it, I could annotate the data with a guess about OS and browser from the user agent string. Cool idea. Thanks!

Re: 90% of the T Distribution

#16

Post by zamadatix »

kqr wrote: post #15
ranma42 wrote: post #14

Earlier quoted context omitted.

How do you distinguish between input (Keyboard/USB) and output (Browser/Graphics stack) delays in this test?

That's outside the scope of what I wanted to find out so my plan was to treat that as part of the random effects varying from user to user. But now that you say it, I could annotate the data with a guess about OS and browser from the user agent string. Cool idea. Thanks!

It won't be very random though. E.g. Apple keyboards will almost always be on device with Apple's hardware & stack, cheap keyboards will tend to be with cheap computers/monitors, expensive keyboards the opposite.

It'll still be interesting data (and I'll dump as many of my systems in as I can), but it will take a lot more than treating differences as noise to answer those kinds of questions in a meaningful way.

Re: 90% of the T Distribution

#17

Post by kqr »

zamadatix wrote: post #16
kqr wrote: post #15

Earlier quoted context omitted.

That's outside the scope of what I wanted to find out so my plan was to treat that as part of the random effects varying from user to user. But now that you say it, I could annotate the data with a guess about OS and browser from the user agent string. Cool idea. Thanks!

It won't be very random though. E.g. Apple keyboards will almost always be on device with Apple's hardware & stack, cheap keyboards will tend to be with cheap computers/monitors, expensive keyboards the opposite. It'll still be interesting data (and I'll dump as many of my systems in as I can), but it will take a lot more than treating differences as noise to answer those kinds of questions in a meaningful way.

That is indeed an annoyance. Last time I looked at the data the "Apple keyboard" and "Apple computer" submissions overlapped so much I couldn't include them separately in the model.

Similarly, I can't separate the effects of programmable and split keyboards because those two almost perfectly overlap in the data.

I hope getting more submissions will help at least a little bit with this.

Re: 90% of the T Distribution

#18

Post by srean »

kqr wrote: post #17
zamadatix wrote: post #16

Earlier quoted context omitted.

It won't be very random though. E.g. Apple keyboards will almost always be on device with Apple's hardware & stack, cheap keyboards will tend to be with cheap computers/monitors, expensive keyboards the opposite. It'll still be interesting data (and I'll dump as many of my systems in as I can), but it will take a lot more than treating differences as noise to answer those kinds of questions in a meaningful way.

That is indeed an annoyance. Last time I looked at the data the "Apple keyboard" and "Apple computer" submissions overlapped so much I couldn't include them separately in the model. Similarly, I can't separate the effects of programmable and split keyboards because those two almost perfectly overlap in the data. I hope getting more submissions will help at least a little bit with this.

It might help to pose the problem as regression analysis with with categorical dependent variables. Perhaps with generalized linear model to account for the case that the predicted variable is positive.

Alternatively, if you have enough data, see if an orthogonal array design is feasible. It will not be very kosher because you would be selecting as opposed to assigning.

Re: 90% of the T Distribution

#19

Post by kqr »

srean wrote: post #18
kqr wrote: post #17

Earlier quoted context omitted.

That is indeed an annoyance. Last time I looked at the data the "Apple keyboard" and "Apple computer" submissions overlapped so much I couldn't include them separately in the model. Similarly, I can't separate the effects of programmable and split keyboards because those two almost perfectly overlap in the data. I hope getting more submissions will help at least a little bit with this.

It might help to pose the problem as regression analysis with with categorical dependent variables. Perhaps with generalized linear model to account for the case that the predicted variable is positive. Alternatively, if you have enough data, see if an orthogonal array design is feasible. It will not be very kosher because you would be selecting as opposed to assigning.

My current best attempt is a mixed-effects quantile regression (to capture the influence of the fixed effects on the tenth percentile while accounting for dependence between trials from the same person) but it is so compute hungry with this (apparently relatively large) data set that I'm looking into Bayesian methods for accomplishing similar things.

Re: 90% of the T Distribution

#20

Post by srean »

kqr wrote: post #19
srean wrote: post #18

Earlier quoted context omitted.

It might help to pose the problem as regression analysis with with categorical dependent variables. Perhaps with generalized linear model to account for the case that the predicted variable is positive. Alternatively, if you have enough data, see if an orthogonal array design is feasible. It will not be very kosher because you would be selecting as opposed to assigning.

My current best attempt is a mixed-effects quantile regression (to capture the influence of the fixed effects on the tenth percentile while accounting for dependence between trials from the same person) but it is so compute hungry with this (apparently relatively large) data set that I'm looking into Bayesian methods for accomplishing similar things.

Quantile regression is such an under-used tool. It really needs to be better known.
Post reply on HN