|
asdf32 posted:You and your null hypothesis. The choice of a null hypothesis is a hypothesis buddy. "No price increases" isn't the clear null hypothesis in the case of a labor cost increase. Is this a joke? Am I living in real life right now? The null hypothesis in a cause and effect study like this one is that there is no relationship between two measured variables. In this case, the null hypothesis is that a change in minimum wage does not cause a change in prices. In order to claim that prices change as a function of the minimum wage, a statistical model describing these variables must reject the null hypothesis. Failing to do that means that any observed relationship between the two variables is a fluke. "No price increase after a 10% minimum wage increase" is the clear null hypothesis. That would be up for debate in a Statistics 101 classroom, perhaps, but not in the real world. quote:Also the statement is incorrect. Assuming identical base data higher level city data isn't necessarily more certain (obviously). Far from it, higher level city data strips out the details of individual restaurants which was a key aspect of this study. Both for looking for mosospony firms within the statistics (in the section you didn't like) and for showing how the competitive model predicted the difference between limited and full service restaurants (which other models did not) By definition, its fractional certainty is greater than or equal to the fractional certainty of the lower level data. I understand the importance of looking at these details for this study, but that does not improve the uncertainties of these measurements. The BLS CPI data is all collected in approximately the same way, and in this kind of case you aren't able to improve your uncertainties by removing measurements. Your argument would only be accurate if there was reason to believe that a large subset of the BLS data was subject to significantly greater uncertainties than the BLS data used in this study, but there is no reason to suspect this. quote:Ok I did miss that row. It figures that you're trying to reject the study based on R squared alone which is a limited metric. If you're trying to claim the results are statistically insignificant you're incorrect. You don't actually know what statistical significance means! I don't want to reject the study based on R squared alone. I want to reject the study based on their inability to reject the null hypothesis. Their extremely low R^2 merely means that their model does not describe the variance in the data well at all. Their model could in fact be a very good fit to the data despite that (because R^2 is not a goodness of fit metric), but that's beside the point; if the variance in the data is so large that their R^2 is basically zero, then it means that their uncertainties are likely enormous, which explains why they didn't bother trying to reject the null hypothesis. But hey, that's the kind of thing to expect from a working paper. I don't fault them for having an extremely low R^2. I fault them for using a statistical model to claim correlation between two variables when that model fails an extremely important litmus test. The low R^2 really just reinforces this point, it's not important on its own.
|
#
?
Jun 22, 2015 04:39
|
|
|
|
| # ? May 15, 2024 18:30 |
|
|
I have a question about the idea of a $15 minimum wage. What happens to people who aren't worth $15 per hour? There are a lot of people, probably millions of people in the U.S., who are marginally worth keeping on a job for $7.50 or $8 per hour, or who really aren't fit for a minimum wage job but manage to go from job to job every few weeks or months. My family owns two restaurants, and many if not most of our employees have lots of serious problems which would make any employer think twice about hiring or keeping them. It's a regular occurrence for an employee not to show up to work because they're in jail again. Two of our managers are alcoholics, and I mean drinking alcoholics, who still manage to do a reasonable job most of the time, because they wait to drink till they're off work. Of course, sometimes they don't show up for work because they started drinking at the wrong time or are too badly hung over to be able to work. Many of our employees are dull witted and slow moving and we're waiting for an opportunity to get rid of them and try to bring on someone better, but haven't gotten to it yet. They can still fill a slot in the schedule, although it takes them two or three times as long as an normal person to do the job. They're not mentally retarded, they're just dumb and not good at anything. What happens to these people if the minimum wage is $15 per hour? There are absolutely people who aren't worth 8.
|
|
#
?
Jun 22, 2015 05:17
|
|
|
Orange Sunshine posted:I have a question about the idea of a $15 minimum wage. What happens to people who aren't worth $15 per hour? Why do you keep employing them for any amount of money? These are clearly people who should be in some kind of rehab, not slaving away for the pittance you deign to provide them.
|
|
#
?
Jun 22, 2015 05:21
|
|
|
Orange Sunshine posted:What happens to these people if the minimum wage is $15 per hour? There are absolutely people who aren't worth 8. Then they're already unemployed.
|
|
#
?
Jun 22, 2015 05:24
|
|
|
Support more expansive welfare systems then. The point of the minimum wage is to provide adequate compensation for those who do work, if you think they aren't worth that then you are dumb for hiring them. Maybe you can't find better employees because you are offering lovely wages and don't want to fork out for quality.
|
|
#
?
Jun 22, 2015 05:32
|
|
|
Orange Sunshine posted:Many of our employees are dull witted and slow moving and we're waiting for an opportunity to get rid of them and try to bring on someone better, but haven't gotten to it yet. They can still fill a slot in the schedule, although it takes them two or three times as long as an normal person to do the job. They're not mentally retarded, they're just dumb and not good at anything. You'll keep them on at 15 an hour, or you'll fire them until you realize you can't run your business without adequate staffing and take the hit elsewhere. Your restaurant may fail, and it will be because your business model depends on a labor cost that only exists due to federal subsidy and welfare. Considering how you talk about your employees, I doubt anyone will shed a tear. Of course, you may figure out how to train and motivate your workforce, but I'm not holding my breath.
|
|
#
?
Jun 22, 2015 07:09
|
|
|
Dr. Arbitrary posted:Can someone with a little more understanding of economic indicators help me with this article from Heritage? It's just a bunch of irrelevant handwaving, I don't think there's much to get. They claim that using different measures of inflation makes the "liberal" graph inaccurate. They go on to show the variance by going back and using different indicators to adjust inflation, but conveniently fail to show how that affects the conclusion in the original graph, i.e. productivity has approximately doubled while the purchasing power of minimum wage has stayed approximately the same. Then they show a graph that proves that average wage matches gains in productivity pretty evenly over the same time period. Which is true, and has nothing to do with the minimum wage, so I have no idea what that's about. As a side note, it's a pretty awful graph, because it fails to include the median wage line and thus glosses over the wage disparity trend that's central to the discussion. Then they show that productivity matches wage pretty closely in the fast food sector. Which is true, or not true, I don't know because they don't cite sources or explain where their numbers come from. Either way, it's irrelevant because the original graph is about the entire economy, not just one sector, which is the point. Then they say stuff like "Neither has productivity [risen] in minimum-wage jobs. Of course, most workers do not stay in low-productivity minimum-wage jobs very long." which are actually just good old fashioned, straight up lies. jesus christ why did i read that whole thing
|
|
#
?
Jun 22, 2015 07:30
|
|
|
Dr. Fishopolis posted:Then they show a graph that proves that average wage matches gains in productivity pretty evenly over the same time period. Which is true, and has nothing to do with the minimum wage, so I have no idea what that's about. As a side note, it's a pretty awful graph, because it fails to include the median wage line and thus glosses over the wage disparity trend that's central to the discussion. I think I get it. I assume that there's nothing surprising that average wage and productivity increase at the same rate. So, average wages is a deceptive measure because theoretically, a chart where the productivity gains were distributed equally and a chart where the productivity gains went entirely to one person would be identical because an average is simply all the income added up and divided by the number of people. I found this completely unbiased chart: http://www.heritage.org/federalbudget/top10-percent-income-earners  It shows that almost half of the income earned is earned by the top 10%. Would this be like measuring the average weight of a group of people that consisted of 9 children and one obese person over time?
|
|
#
?
Jun 22, 2015 07:52
|
|
|
QuarkJets posted:Is this a joke? Am I living in real life right now? #1 more detailed data is better in this case. The noise that cancels out under aggregation also cancels out in OLS. Restaurant level data is much better than CPI data because you have more variation in both x and y #2 ofc r2 is low in a study like this, price variation is largely independent of the minimum wage because minimum wages are fairly constant and only affect a very small portion of workers. The significance and direction of the coefficient are all that matter. You get a similar r2 in studies between smoking and life expectancy... You're explaining a continuous variable with significant outside variation with a dummy like variable.
|
|
#
?
Jun 22, 2015 10:21
|
|
|
Geriatric Pirate posted:Jesus this is what happens when someone who took stats 101 tries to sound smart.. I know I'm wasting my time because you're just going to make up stuff in response but whatever I've actually been using statistics in a professional context for over a decade. I'm not here to browbeat you with credentials, but don't fool yourself into believing that my experience is limited to a single undergraduate course. quote:#1 more detailed data is better in this case. The noise that cancels out under aggregation also cancels out in OLS. Restaurant level data is much better than CPI data because you have more variation in both x and y There are a number of reasons why this is not true. This is an issue that we frequently encounter in high energy physics; at best you can try to claim that the uncertainties are the same with the data subset, but usually they're worse for a number of reasons (for example, sometimes systematic uncertainties that are important to smaller datasets get ignored because someone analyzing a much larger dataset chose to ignore them). Regardless, the uncertainties definitely aren't any better. You could try to make the argument that they're the same (unlikely), but that still leaves you up poo poo creek without a paddle. quote:#2 ofc r2 is low in a study like this, price variation is largely independent of the minimum wage because minimum wages are fairly constant and only affect a very small portion of workers. The significance and direction of the coefficient are all that matter. You get a similar r2 in studies between smoking and life expectancy... You're explaining a continuous variable with significant outside variation with a dummy like variable. Did you just skim my post or did you read it and just not comprehend it? We agree that the R^2 is not really an important quantity. We already know that the data has enormous variance, so it's not surprising that the R^2 is low: QuarkJets posted:...Their model could in fact be a very good fit to the data despite that (because R^2 is not a goodness of fit metric)... This study's problem is not its low R^2.
|
|
#
?
Jun 22, 2015 11:39
|
|
|
QuarkJets posted:I've actually been using statistics in a professional context for over a decade. I'm not here to browbeat you with credentials, but don't fool yourself into believing that my experience is limited to a single undergraduate course. . It's not difficult. Their dataset is much better than the aggregate data. Re r2 then why bring it up if even you know it's not relevant in a case like this where you just want to test one coefficient instead of fitting model?
|
|
#
?
Jun 22, 2015 12:15
|
|
|
Ardennes posted:Your not responding to my point. If that's what you got from reading the numbers you should go back and consider both the theoretical model they put forward to explain how much is passed through, and how they went about assessing whether the minimum wage was an effective floor. You're quoting numbers without understanding where they came from or what they mean. Which is not an uncommon phenomenon in this thread, unfortunately. *reads subsequent couple of pages* Yeah.. =/ Absurd Alhazred posted:Why do you keep employing them for any amount of money? These are clearly people who should be in some kind of rehab, not slaving away for the pittance you deign to provide them. It's pretty hard to find people who don't have some kind of issue, even for more highly paid positions. A whole lot more people are functional wrecks who can just sort of hold it together to keep a job than you would imagine. So as an employer, sometimes you'd rather work with the team you've got and work around their known deficiencies than force turnover and have to manage a new team with all new problems. I can totally understand where Orange Sunshine is coming from.
|
|
#
?
Jun 22, 2015 14:35
|
|
|
wateroverfire posted:It's pretty hard to find people who don't have some kind of issue, even for more highly paid positions. A whole lot more people are functional wrecks who can just sort of hold it together to keep a job than you would imagine. So as an employer, sometimes you'd rather work with the team you've got and work around their known deficiencies than force turnover and have to manage a new team with all new problems. I can totally understand where Orange Sunshine is coming from. Empathy? Who are you and what have you done with wateroverfire?
|
|
#
?
Jun 22, 2015 14:36
|
|
|
Geriatric Pirate posted:Re: Your second paragraph, did you just decide to throw in as much jargon as you could fit into one paragraph? It doesn't mak any sense. I'm not a physicist but my non-swamp related education has lead me to believe that he is talking about physics on an atomic or subatomic scale which is pretty much all probability related. quote:Re r2 then why bring it up if even you know it's not relevant in a case like this where you just want to test one coefficient instead of fitting model? Because it reinforces the data as being worthless, just because a piece of supporting evidence isn't strong enough on its own it can still provide heft to a point.
|
|
#
?
Jun 22, 2015 16:52
|
|
|
Raskolnikov38 posted:I'm not a physicist but my non-swamp related education has lead me to believe that he is talking about physics on an atomic or subatomic scale which is pretty much all probability related. It doesn't provide any heft because it's an irrelevant point, just like saying the sky is blue 10 times after arguing that MMT isn't crazy heterodox fringe crap doesn't make the original statement any more true
|
|
#
?
Jun 22, 2015 17:06
|
|
|
Dr. Arbitrary posted:I found this completely unbiased chart: yes, also holy poo poo that graph. it's like tracking the weight of 100,000 mayflies relative to a single elephant, and giving the mayflies one single color bar and the elephant the other. Most humans would come away surprised at how heavy mayflies really are.
|
|
#
?
Jun 22, 2015 17:27
|
|
|
Geriatric Pirate posted:It doesn't provide any heft because it's an irrelevant point, just like saying the sky is blue 10 times after arguing that MMT isn't crazy heterodox fringe crap doesn't make the original statement any more true Hey that reminds me I'm still waiting for anything from MMT saying infinity debt is good.
|
|
#
?
Jun 22, 2015 17:34
|
|
|
Raskolnikov38 posted:I'm not a physicist but my non-swamp related education has lead me to believe that he is talking about physics on an atomic or subatomic scale which is pretty much all probability related. While the theory of quantum mechanics does say that things are intrinsically probabilistic, and so physicists do use and have some education in probability, if we are playing the game of what disciplines do we appeal to authority on the subject of statistics, I would say that the social scientists use it more often, and at least should be more fluent. Although a lot of social scientists are pretty bad at math. Physicists often tend to have the belief that they are the best at everything and are better than biologists at biology, mathematicians at math, computer scientists at computer science, and so on. Statistics isn't really as important of a part of a physics education as it is in the social sciences. That's because physicists bend over backwards to study the simplest systems and since the theory is better for the simple systems, they don't have to resort to statistics as often to understand them. You tend to lean on statistics more heavily when you have no idea what is going on and need to do trial and error experiments to try to figure stuff out.
|
|
#
?
Jun 22, 2015 17:40
|
|
silence_kit posted:While the theory of quantum mechanics does say that things are intrinsically probabilistic, and so physicists do use and have some education in probability, if we are playing the game of what disciplines do we appeal to authority on the subject of statistics, I would say that the social scientists use it more often, and at least should be more fluent. Although a lot of social scientists are pretty bad at math. Physicists often tend to have the belief that they are the best at everything and are better than biologists at biology, mathematicians at math, computer scientists at computer science, and so on. I would say that statisticians are probably the best authority on statistics, and when it comes to elementary statistics like R^2, a physics education is going to provide plenty of training in the form of lab courses.
|
|
|
#
?
Jun 22, 2015 17:48
|
|
|
silence_kit posted:While the theory of quantum mechanics does say that things are intrinsically probabilistic, and so physicists do use and have some education in probability, if we are playing the game of what disciplines do we appeal to authority on the subject of statistics, I would say that the social scientists use it more often, and at least should be more fluent. Although a lot of social scientists are pretty bad at math. Physicists often tend to have the belief that they are the best at everything and are better than biologists at biology, mathematicians at math, computer scientists at computer science, and so on. It might not be a huge part of your core physics curriculum (BS/MS), though it is a part. But if you're talking about using huge quantities of data to measure very small effects, you are going to get familiar with statistics pretty quick. Today's high energy physics is very much in this realm, as are things like trying to check if certain quantities (like the electron electric dipole moment) are zero or just very close to zero.
|
|
#
?
Jun 22, 2015 18:14
|
|
|
Akumu posted:It might not be a huge part of your core physics curriculum (BS/MS), though it is a part. But if you're talking about using huge quantities of data to measure very small effects, you are going to get familiar with statistics pretty quick. Today's high energy physics is very much in this realm, as are things like trying to check if certain quantities (like the electron electric dipole moment) are zero or just very close to zero. Haha, ok, ok fair enough. Some physicists do spend their entire careers trying to measure astonishingly weak effects, so I am willing to believe that they would become statistics experts so that they could continue to study stuff like that.
|
|
#
?
Jun 22, 2015 18:19
|
|
|
Geriatric Pirate posted:Jesus this is what happens when someone who took stats 101 tries to sound smart.. I know I'm wasting my time because you're just going to make up stuff in response but whatever I like that you jumped right ot this instead of actually dealing with the multiple people calling you out for saying the working poor don't exist, as opposed to just calling them liars for sharing their experiences being working poor.
|
|
#
?
Jun 22, 2015 18:31
|
|
|
Geriatric Pirate posted:Re: Your second paragraph, did you just decide to throw in as much jargon as you could fit into one paragraph? It doesn't mak any sense. You can look at restaurant level data or aggregate the data at a city level. If there's unbiased measurement error it's not a problem and you can deal with it even with ols (most times), if there's bias then aggregation is also biased Hahahahahahahahahahahahahaha Are you serious? How many of the words in that paragraph do you think are jargon? Seriously, please list all of the jargon terms. This is such a rich twist, in the previous post you accused me of having no real statistics knowledge, and now you accuse me of using too much statistics jargon because I whipped out a big word like "systematic", perhaps? If there are unbiased fractional uncertainties, they're not going to be decreased by decreasing your sample size. Same for biased fractional uncertainties. This is such an elementary point of statistics that I don't really know how else to drive this point home. You keep trying to claim that the smaller dataset has smaller fractional uncertainty, but it really, really doesn't. OLS is also being used by the other papers, and they had issues because the uncertainties were too large, so this paper should have at least the same issues! There are cases where this could be false, but you haven't given a good reason for why that would be the case. In your world it would be possible to drive down uncertainties by just randomly selecting a subset of data, but that is clearly not the case You keep trying to use OLS as an excuse to ignore the large uncertainties, but surely you realize that OLS doesn't actually make those uncertainties go away, right? OLS can help drive down the final propagated uncertainty on your reported values, but those original uncertainties are still going to bite you in the rear end when you attempt to reject the null hypothesis, which is why this paper failed to do that. quote:Re r2 then why bring it up if even you know it's not relevant in a case like this where you just want to test one coefficient instead of fitting model? It's a clever trap to lure idiots like you into revealing that they're illiterate The R^2 is relevant, it's just not as important as the fact that the authors couldn't reject the null hypothesis. It's a supporting detail. The small R^2 supports the notion that the data has very large variance, which implies very large uncertainty. We know that both of these things are true, and the small R^2 helps to confirm it. It's like finding a bullet casing at the scene of a shooting; we already know that a gun was used, the casing is just a supporting detail (not like whatever dumb bullshit "sky is blue" thing you tried to use in a later post, god drat you suck at analogies) You should try actually read the things that you're responding to before you respond to them.
|
|
#
?
Jun 22, 2015 18:44
|
|
|
Thread, am I wrong? Did I actually vomit out a bunch of jargon terms that I assumed would be commonly understood? Are terms like "dataset", "subset", and "uncertainty" actually jargon terms? I would assume that you don't need any statistics knowledge to understand what these words mean. The only word that I can find in that entire second paragraph that could be considered jargony is "systematic", are there others?
QuarkJets fucked around with this message at 18:52 on Jun 22, 2015 |
|
#
?
Jun 22, 2015 18:49
|
|
|
ITT we learn that physicists are definitely better economists than economists. In other news, to no one's surprise the thread remains lovely.
|
|
#
?
Jun 22, 2015 18:49
|
|
|
QuarkJets posted:Thread, am I wrong? Did I actually vomit out a bunch of jargon terms that I assumed would be commonly understood? Are terms like "dataset", "subset", and "uncertainty" actually jargon terms? You are absolutely sperging out over some irrelevant and meaningless bullshit, yeah. You're down so deep in the weeds talking about data you haven't examined and a study you probably haven't more than glanced at in a field that isn't yours that the entire critique is ridiculous on a level verging on a self parody of physicists so brilliant that I pray it's intentional.
|
|
#
?
Jun 22, 2015 18:54
|
|
|
wateroverfire posted:ITT we learn that physicists are definitely better economists than economists. I don't think so. I'm not able to build a model with economic data or anything like that. But anyone with a statistics background should be able to recognize the importance of rejecting the null hypothesis before drawing conclusions from your model, right?
|
|
#
?
Jun 22, 2015 18:54
|
|
|
wateroverfire posted:ITT we learn that physicists are definitely better economists than economists. Here's a whole list of economists who are worse economists than a 7 year old child
|
|
#
?
Jun 22, 2015 18:55
|
|
|
haha i love that this really dumb guy is still posting because he's stubborn about being wrong
|
|
#
?
Jun 22, 2015 18:55
|
|
|
wateroverfire posted:You are absolutely sperging out over some irrelevant and meaningless bullshit, yeah. You're down so deep in the weeds talking about data you haven't examined and a study you probably haven't more than glanced at in a field that isn't yours that the entire critique is ridiculous on a level verging on a self parody of physicists so brilliant that I pray it's intentional. Failing to reject the null hypothesis is irrelevant? Can you tell me why?
|
|
#
?
Jun 22, 2015 18:55
|
|
|
You know, it's really creepy to see people repeat the exact same "well you're not trained in this field" argument simultaneously. Are we sure D&D's right wing are human, and not the puppets of a monstrous and alien intellect?
|
|
|
#
?
Jun 22, 2015 18:56
|
|
|
Orange Sunshine posted:I have a question about the idea of a $15 minimum wage. What happens to people who aren't worth $15 per hour? but what about all the people who aren't worth the desperation wages i piteously give them 
|
|
#
?
Jun 22, 2015 18:57
|
|
|
QuarkJets posted:Failing to reject the null hypothesis is irrelevant? Can you tell me why? Back up to what the hypothesis was and why I should care about this line of questioning and I'll get back to you?
|
|
#
?
Jun 22, 2015 18:58
|
|
wateroverfire posted:Back up to what the hypothesis was and why I should care about this line of questioning and I'll get back to you? The null hypothesis is that there is no relationship between two measured phenomena. In order for a hypothesis to be meaningful, it must beat the null hypothesis. This is basic science. The inability of a hypothesis to beat the null hypothesis, statistically, suggests that the hypothesis is bullshit.
|
|
|
#
?
Jun 22, 2015 19:01
|
|
|
look, for me to give a poo poo about your claim that my stats are bullshit you're going to have to explain this "standard deviation" if you didn't just make that up 
|
|
#
?
Jun 22, 2015 19:03
|
|
|
Popular Thug Drink posted:look, for me to give a poo poo about your claim that my stats are bullshit you're going to have to explain this "standard deviation" if you didn't just make that up Alternately, "direct your methodological sperging at the Chicago fed if you think you have a point", I guess. Also maybe read the paper.
|
|
#
?
Jun 22, 2015 19:07
|
|
|
hell, i've taken econ 1. 0. 1. don't even step to me, boy
|
|
#
?
Jun 22, 2015 19:08
|
|
|
You people! We can't autism about methodology, there's science to do!
|
|
|
#
?
Jun 22, 2015 19:09
|
|
|
Popular Thug Drink posted:hell, i've taken econ 1. 0. 1. don't even step to me, boy Maybe read the paper, IMO.
|
|
#
?
Jun 22, 2015 19:12
|
|
|
|
| # ? May 15, 2024 18:30 |
|
|
wateroverfire posted:Back up to what the hypothesis was and why I should care about this line of questioning and I'll get back to you? The null hypothesis is that there's no relationship between the price data and the unemployment wage. Failing to reject the null hypothesis means that the conclusions from the paper may as well be fart gas. It means that the authors may as well have said "we observed no significant effects". You just said that this isn't important. Can you explain why?
|
|
#
?
Jun 22, 2015 19:16
|
|