Recently doing HW with a T/F question:
> In a multiple regression model, regardless of the shape of population distribution for an independent variable, the sampling distribution of a regression slope is normally distributed
I of course responded true and evoked the all powerful CLT. Now I got points off with the following comment:
> good, but wanting to have some indication of what impact the predictor distribution has -- does it impact the residual distribution?
I followed up with the following:
> After some further research and reflection I am not 100% confident in what I think. My understanding is that we assume the error distribution in our model, but residuals (used to estimate error) are a function of y and y-hat, where y-hat is E(y_i | x_i) (ie integrating over conditional PDF of Y given the predictor(s)). Therefore, in some ways the predictor distribution is linked through fitted values. However, with OLS there is no explicit assumption of normality, it just so happens that normal MLE and OLS converge, but we can generalize our model and assume different distributions.
> I have also seen CLT argument such as:
> "rationale for normally distributed errors is often phrased in terms of the many factors influencing y being additive, appealing to the Central Limit Theorem to suggest that the sum of a large number of random factors will be normally distributed."
> But I hesitate to accept this because such arguments are controversial since people like to loosely employ CLT/asymptotic arguments which can at times be called “unfounded". Furthermore, it seems to me that the kind of data you are modeling affects normality of errors or more accurately whether it is reasonable to assume normal errors. In conclusion, it is difficult for me to give a conclusive answer.
The response to this:
> True, the distribution for X can be anything (think about binary predictors, etc.) and the sampling distribution for the regression coefficient depends on the distribution of the residuals--and even if these are modestly non-normal, the sampling distribution will be normal--especially as the sample size increases. By the way, your reasoning in your email was spot on.
Now I am confused for the following reasons:
- The original question was pertaining to the hypothetical effect of predictor(s) distribution on the residual distribution, but the response now flipped it back to focusing on the sampling distribution and its dependence on the res
Hello everyone,
I would very much appreciate if someone could help me clarify a concept I have been struggling with. So basically im trying to uderstand the difference between errors are residuals.
How I currently understand it is that errors are unobserved and residuals are merely a by product of estimation. Hence, its like saying residuals are NOT the sample counter part of errors. All that is fine. However, we try to consistently estimate the variance of the error term in many of our asymptotic derivation. We, of course end up having some variance covariance matrix which depends on the variance of the error term. However, what confuses me is that we estimate this using the residuals (which kinda goes against the logic of them not being the sample counter part of the errors). In the proof of estimation we cannot rely on the weak law of large numbers (precisely i think because of the argument above) to say that the sample average of squared residuals consistently estimate the errors variance. Instead what we do is purely algebraic (false 0 and bounding things in probability so that they are little ohh pee one).
So whats the connection?
Any help is greatly appreciated!
Pretty interesting. Definitely think the Butler trade was the right risk to take, but Covington and Saric seemed to play a large role in Simmons’ effectiveness. Those two players combining for half of Simmons 3 point assists feels like a lot. Who replaces that production in the short term? Who can they grab in the long term?
What is the difference and when should i do either or,for most efficient workflow.
Edit: Thank you all so much for your input and suggestions really helped me optimize my work flow when it comes Lows and holes and residual scans.
Hey guys- yesterday I got my results from my first post RAI bloodwork. It looks like I still have residual cancer and it’s still growing even after my RAI last February. I’m pretty freaked out. My endocrinologist wants me to take an extra half pill of my levothyroxine to suppress tsh and then recheck in 3 months but I’m kind of impatient and I just want to see what we can do to actually get RID of this stupid cancer. I’m only 32 and I feel like this cancer just keeps my life on hold indefinitely. My husband and I want to have kids and we finally found a possible solution to one of the problems on his end, but then this result comes up and I feel like I can’t move forward with my life.
Has anyone else here had persistent papillary thyroid cancer? What did you do??
Thanks
Hey all. Sort of a check-in post here for my sake, plus I felt like writing. Maybe someone will get something out of it, too.
It's been a years-long path but recent developments in my journey have really made me start to trust in myself more. I'm not being so hard on myself anymore. I had a sort of deep experience last week that wasn't the first, but was important to me. In short I've really been opening myself up to someone these last few months, little by little. I can't exactly explain it but this sort of thing has backfired on me in the past. I guess I've made mistakes with it. But last week I got confirmation that everything is going OK, better than OK, actually... everything is going great between us.
So I realized all that posturing and trying to "hang on" was unecessary. I don't need to say the right things, do the right things, be constantly on guard... it's OK to just be myself.
Not that I'm saying it's always OK or good to open up to everyone... it's not. That's the root of the mistakes I've made with this stuff... opening up in inappropriate ways at inappropriate times. But the overall trajectory has been correct. I'm learning.
I guess I was letting my mistakes confirm this idea I had about myself that "I'm a fuck up". But it's a lie. I'm just a human.. we all make mistakes... especially when it comes to uncharted life experiences.
And I think deep down I wasn't even afraid of getting it wrong. I think I was afraid of getting it right, you know? Cus opening myself up opens me up to a new experience. It's scary. So in a way it's easier to just say "I'm a fuck up" and crawl back into my hole. I'm glad I didn't.
So a big part of this self-punishment and need to "get things right" has dropped. I'm much more comfortable just being me. I can see it in a deep way, but even in the small things. I live in a foreign country and, though I'm fluent in the language, I used to get so frustrated about little mistakes I would make. I'm not doing that as much anymore. It's funny.
The second part of this is that I'm letting go of control in general. The coronavirus situation has brought all sorts of changes suddenly. There's a lot of uncertainty, and every day is something new. There's really no routine anymore. Well, I've just been going with the flow. I'm staying up quite late, but I'm still getting 6 hours of sleep, plus a 2-3 hour nap later in the afternoon. I don't know. I just don't know what to expect anymore.
... keep reading on reddit ➡Example of residuals in question
When writing a paper, of course we should list the convergence of our simulations, but I'm unsure of which form of residuals I should list. In STAR-CCM+, the default is to normalize the residuals by the maximum residual in each cell from the first 5 iterations, so that'll change based on your initial solution, and probably on things like whether you've initialized from zero velocity, freestream velocity, or used grid sequencing initialization as well. This is good to gauge relative convergence, because all of your residuals will start at ~1 in the first few iterations, so it's easy to tell how much they've dropped initially.
My gut feeling tells me this is wrong to write in a paper, because it's all relative and doesn't give an absolute reading on your solution. But if I don't normalize, I get a residual >1E-3 on my specific dissipation rate, which seems high compared to "academic quality" results. Especially when I'm trying to replicate a paper getting 2 orders of magnitude better convergence with 30% less cells, using the exact same geometry, physics, and software.
Meanwhile you can see from the normalized version that Sdr dropped 7 orders of magnitude from the initial solution! Absolute convergence? Kinda. Relative convergence? For sure.
To normalize or not?
I won’t need to say a magic word, physically move my joints, or perform an act to use the ability; rather, I will simply think it and it will be done (with intent of course).
The ability would also have a cool down where it can’t be activated again until all the dopamine from the pleasure of the previous pops dissipates. The pleasure from popping my joints will never diminish.
Furthermore I can bestow this ability on others if wished. I can also use it on others while being still being able to adjust the levels of the conditions stated in the title.
It also includes the condition of sounding like falling rain or crackling leaves if I know I’m in front of someone grossed out by the sound if I choose to be kind.
Hello everyone,
I'm doing a counter flow heat transfer problem in star ccm. I've got everything but Y+. This is a Y+ plot from my simulation and as you can see the scale is from 0.01 to 5.9429 but all Y+ along the tube are around 0.2. The only high Y+ occurred at the air outlet which is the small pipe downward at bottom left of this pic. I'm using K-E model with V2F low wall Y+ treatment. Residual is not diverging nor converging but fluctuating at 1e0 level. This is very confusing but I think the problem is Y+. According to Y+ equation and with Y+=1, I calculated the first layer of cells have to be 0.095mm in my case. So using growth factor of 1.2 I calculated the prim layer thickness for 5 layers is about 1mm. So I used this setting to mesh my prism layer. I'm not sure if I'm doing this right. Can someone help me with this issue? Thank you all very much.
https://preview.redd.it/yjla7mjvrzc31.png?width=702&format=png&auto=webp&s=eeda61acfd59b79378754a30266d62ef1da20d11
Ended up in hospital on my first time doing 0.7-1.0g of questionable amphets. Edit: have not slept in nearly 2 days also.
Had this happen to me just now. I used a ton of soap, and it would not come off. Ended up googling it, trying vegetable oil, and it worked like witchcraft.
"Monty Python and the Quest for the Holy Grail" / "Monty Python and the Holy Grail"
Anyone with nerdish leanings eventually gets exposed to Monty Python. The pinnacle of that experience, and a cornerstone of my youth was, "Monty Python and the Quest for the Holy Grail".
Screen 1: \"Monty Python & the Quest for the Holy Grail\"
I can recite most of the movie by heart, and have seen it conservatively 20+ times, but amazingly, I got the title "wrong". Apparently, it's only ever been "Monty Python and the Holy Grail". It's missing half the title now! Even confirming on IMDB, the "Quest" part has never been a working or alternative title.
There's residuals in print, from movie listings to reviews, that list the film with "Quest". Checking on Twitter, and hundreds of people are discussing, "Monty Python and the Quest for the Holy Grail", which seems like a very specific "misremembering".
Which do you remember, "Monty Python and the Quest for the Holy Grail" or "Monty Python and the Holy Grail"?
Album of "Monty Python and the Quest for the Holy Grail" residuals:
https://www.flickr.com/photos/nathanielhebert/albums/72157676479740527
That’s my only complaint. Wellick was a really interesting character.
I just rewatched the “Will you be loyal to me?” scene and it’s freaking amazing. “I will ALWAYS be loyal to Elliot.”
