Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Categories
- 32.3K All Categories
- 27.2K LSAT
- 16.5K General
- 5.1K Logical Reasoning
- 1.3K Reading Comprehension
- 1.7K Logic Games
- 72 Podcasts
- 193 Webinars
- 9 Scholarships
- 193 Test Center Reviews
- 2K Study Groups
- 108 Study Guides/Cheat Sheets
- 2.4K Specific LSAT Dates
- 13 June 2024 LSAT
- 3 April 2024 LSAT
- 11 February 2024 LSAT
- 23 January 2024 LSAT
- 37 November 2023 LSAT
- 42 October 2023 LSAT
- 12 September 2023 LSAT
- 38 August 2023 LSAT
- 27 June 2023 LSAT
- 4 August 2024 LSAT
- 30 Sage Advice
- 4.8K Not LSAT
- 4K Law School Admissions
- 11 Law School Explained
- 10 Forum Rules
- 581 Technical Problems
- 276 Off-topic
Related Discussions
Law School Predictor
ChiTownGuy
Alum Member
http://www.lawschoolpredictor.com/wp-content/uploads/Law-School-Predictor-Full-Time-Programs.htm
What are the chances JY and the gang @"Dillon A. Wright" can secure the rights to this nice little tool right here. Given that it hasn't been updated since 2013, I'm sure that the creators wouldn't mind if 7sage buys the rights and updates this and incorporates it into their site. Would be pretty badass if you ask me, and it's a shame to see something so sophisticated sitting off in the corner gathering dust like this, could be tremendously useful to most folks I would think.
This discussion has been closed.
Comments
Ahh I don't think something this simple would be worth investing in. Especially since MyLSN does a decent job at chancing applicants.
@"Alex Divine" Yeah initially those were my thoughts until I actually sat down to create it. Needless to say working with spreadsheets that have an excess of 1 million cells of data (Mylsn data) isn't something that is easily handled on most peoples at home computers. Were you able to create it easily? Mind sending me your model?
Also, I think that the lack of any other similar logistical regression model available on the entire internet speaks to the nature of the difficulty involved in creating one.
I don't think that @_CavsFan is right about the difficulty of making model used by the law school predictor. However, I think it is actually such a small improvement over the data searching tool already available on LSN that aquiring would be almost totally useless.
The LSN data is in a very nice .csv file that is pretty easy to deal with even with my rudimentary programming knowledge. You can download it and/or contact the curator of the data through the email on this link. The data was updated 84 days ago from the main website. http://mylsn.info/download-raw-data/
LSN also has a tool for sorting through the data by LSAT score, GPA range, URM status, early decision, and which cycles of data you would like to include. It simply tells you how many people applied with your selected number range in that year, what percntage got in, what percentage of those that got in reported recieving scholarships, and what the average scholarship ammount was. I don't think that the law school predictor was doing anything all that much more impressive. It's the accumulation of data along with the ability to sort through it that matters here. This search tool is linked to below along with searchable criteria I was using near my present numbers.
http://mylsn.info/3zk4rq/
It's definitely nothing crazy, and I only possess, at best, average programming skills and I'm 90% confident I could tackle this project if I had the motivation/time. What I do for work isn't terribly dissimilar from a project like this. But as @"Seeking Perfection" rightly mentions, the fact that it would be such a small improvement over the free tools on MyLSN makes me think it isn't exactly worth it. This likely is another explanation for why not one else has created something like this; not because it's difficult per say, but because MyLSN kind of does something very similar.
@"Seeking Perfection" @"Alex Divine" The search tool on my LSN is a display of historical data. This predictor is a logistical regression, I think that instead of arguing with you foks I would just refer you to google to understand the difference between those two things.
I also graduated from a top 20 undergrad business program majoring in business stats, and was still unable to create a regression model using the data given it's size and the constraints of cpu processing power processed by y laptop (it just freezes and crashes, its not that I don't know how to do it, my computer won't cooperate its not powerful enough I guess?) I am aware of excel sort functions, I also have worked at a big 4 accounting firm for 4 years now, this was after managing IT operations at a fortune 100 company for 2 years- this is just to give you some indication that I also have worked with excel and am familiar with it. If you don't "think I am correct" please feel free to create such a model and provide one if you are able to easily create I, I would appreciate that.
One of the things about this predictor is that it accounts for URM status in a sophisticated manner, something that makes this tool stand out among any other resource available on the internet.
Also, I would suggest taking at least a brief look at the "about" page on that predictor before making assumptions about the difficulty in creating such a model, it's actually rather involved.
I agree with almost everything you said as well as much of what SeekingPerfection said. I agree it is definitely involved and may take some time to figure out all the intricacies, but once you have the data to pull from, the actual programming and modeling involved wouldn't be that complex. Again, this isn't an assumption, it's just I'm well aware of my competency level WRT modeling and programming. I can think of a few things I've had to do in the last month that are probably more difficult. Further, that's just my opinion based on my background and skills. I probably shouldn't have characterized the project as "simple", when, in fact, it's probably a bit more involved than I would feel comfortable calling simple. I just think it is relatively simple compared to much of the modeling and programming I've done. Sorry if what I originally wrote came across as anything else
Maybe in the future we can team up and put our brains together and build something even better!? Right now, I just don't have the time or motivation to undertake such a project. Plus, I also don't think it's really worth it for the reasons Seeking mentions above.
I don't think I in anyway represented the search tool as not based on historical data or the law school predictor as based on something other than a fairly simplistic regression of that same data.
I don't think the advantage provided by the regression is all that great in part because most of the assumptions which you make as part of such a regression are not close to being satisfied here.
I don't know why you are talking about having worked with excel in the course of obtaining a degree in business. It's rather superfluous to the problem. Excel is notoriously bad at dealing with large sets of data. Even microsoft access would be preferable to Excel, but once more with even my amateurish programming knowledge I would know better than to try to use Microsoft Excel to run a regression here. I hate Microsoft Access so with a small data set I would use Stata or if that was not available R. With a large data set I would either do something using Python(the open source free language I'm familiar with) or talk to one of my CSE friends and have them whip something up using a more appropriate program after I explained the Statistics of it.
But, when I say that I don't think it would be hard, that's on a relative scale. It would certainly take me either a significant amount of time or favors from friends. I cannot justify those expenditures for creating a regression with data which doesn't conform to basic expectations of covariance between values and the error term and will inherently be either biased or extremely wide in its prediction ranges. I would hope 7sage wouldn't consider wasting resourses buying such a program either.
The reason that the tool deals with URM status any better than other predictors is because the underlying data includes more finely detailed information on it, not some imaginary sophistication.
So to summarize, I am confident that you are wrong that the model used by law school predictor is of unusual sophistication, am confident that I could either write a program or easilly find someone who could write a program which would run a similar regression, am confident that the said regression would be fairly flawed, and finally am confident that a solution should not involve the use of microsoft excel.
P.S.
I also care even less about the rank of the place where you went to undergrad(and got a business degree) than the miniscule to non-existent degree which a properly conducted regression would reveal that the law schools care about undergraduate institutions.
@"Alex Divine" Definitely. My intention in making this post was just to indicate that it is a useful tool that would serve a lot of people, and 7sage and it's users would benefit from updating it and incorporating it into the site. And that goes especially for URM's out there (like myself), given the ability of the tool to logically and coherently account for URM status in the admissions probability. @"Dillon A. Wright"
I think that now that we may have cleared up any confusion about the difference between this tool and the myLSN search tool, more folks may be able to understand and appreciate the ability of this tool to provide an alternate means of deriving the statistical probability of admission into law schools!!
@"Seeking Perfection" You need to fall back with that attitude. Whatever you're own personal opinions are on using a regression versus using historical data is irrelevant to me and every one else on this forum. Having a business degree in statistics means that I understand them, which clearly was missed on you. If you can create one so easily than do it.. and idk what "assumptions" your talking about here, that's the point of a regression is to fit a line to a bunch of data points to provide predictions based on variables. Anything else you'd like to get off your chest here? with that long drawn out completely valueless response you've provided other than demonstrating your proficiency in using a bunch of terms you don't understand? It called thinking before you speak (or attempt poorly constructed attacks).
@"Seeking Perfection" Word of advice, trying to argue with someone who's degree was in stats while simultaneusly acknowledging you don't understand stats is a poorly chosen method of argument. And I'm sure in your imaginary world it would be easy to create all sorts of things, but why don't you try entering reality and actually trying to back of some of the clearly outrageous claims you made. Does is matter what tool you use to run regression? No it does not. And given that a top 20 business programs taught that method using .. you guessed it: excel, I would warrant a bit of forethought before attacking people for doing something you don't even understand how to do or use.
"However, I think it is actually such a small improvement over the data searching tool already available on LSN that aquiring would be almost totally useless." Um, small improvement or something totally different? I think the latter is correct. Do you understand that mylsn and this a regression are two different ways of viewing data, each with their own respective qualities? That's something I learned in my studies, as "superfluous" as that may seem to you.
And one more quick thing, I think that the use of excel for 95% of business applications speaks a fair bit to it's appropriateness in modeling data. Have you ever actually had a job where you worked with data? I've done so for the last 6 years. So I understand how you may feel top business programs, fortune 100 companies, and the largest professional services firm in the world might all be wrong about using excel, but based additional off my own extensive expereince I think the use of the tool would be warranted? I'm sure you disagree based off your own tremendous knowledge of the subject. "Amateurish" I think you called it?
@"Seeking Perfection" "However, I think it is actually such a small improvement over the data searching tool already available on LSN that aquiring would be almost totally useless."
"I don't think that the law school predictor was doing anything all that much more impressive."
Above is an indication that you don't understand the difference between myLSN data view and a regression data model.
Below is a contradiction of your own statements. So, if you understand the differences between a regression and historical data view, how is it that you don't "think they are doing anything more impressive"?
"I don't think I in anyway represented the search tool as not based on historical data or the law school predictor as based on something other than a fairly simplistic regression of that same data."
@"Seeking Perfection" "I don't think the advantage provided by the regression is all that great in part because most of the assumptions which you make as part of such a regression are not close to being satisfied here."
Can you be more specific about those assumptions that you are referring to? Can you explain what assumptions those would be?
@"Seeking Perfection""I don't know why you are talking about having worked with excel in the course of obtaining a degree in business. It's rather superfluous to the problem."
This was meant to provide my qualifications for discussing statistical data interpretations. What are your qualifications?
@"Seeking Perfection" "I cannot justify those expenditures for creating a regression with data which doesn't conform to basic expectations of covariance between values and the error term and will inherently be either biased or extremely wide in its prediction ranges."
Can you explain what you meant to convey by that statement? I have a degree in stats and have no idea what "basic expectations of covariance between values and the error term" is referring to.
How would the data be"inherently biased"? can you explain what you mean by that?
As for the tools accuracy, below is taken directly from the site. Do you have any accuracy measures for your predictions to share with us?
"P.S.
I also care even less about the rank of the place where you went to undergrad(and got a business degree) than the miniscule to non-existent degree which a properly conducted regression would reveal that the law schools care about undergraduate institutions."- Would you elaborate on how the below accuracy measures are considered "minuscule to non-existent" or "extremely wide in its prediction ranges"?
Target admit rates, based on how LSP renders predictions:
When LSP said Admit: >= 87%
When LSP said Strong Consider: ~ 69%
When LSP said Consider: ~ 50%
When LSP said Weak Consider: ~ 31%
When LSP said Deny: <= 13%
Based on the 33,500+ LSN decisions from the 2008-09 admission cycle:
When LSP said Admit, 8402 instances: 87.0%, +0.0%
When LSP said Strong Consider, 4842 instances: 78.0%, +9.0%
When LSP said Consider, 10066 instances: 50.6%, +0.6%
When LSP said Weak Consider, 3646 instances: 22.8%, -8.2%
When LSP said Deny, 6754 instances: 10.0%, -3.0%
Difference between actual admit rate and predicted admit rate: (prediction categories weighted evenly)
Average rate: -0.3%
Median rate: -3.0%
Summary:
Prediction categories that were right on the money: Admit, Consider
Prediction categories that were reasonably close: Deny
Prediction categories that were somewhat off: Strong Consider and Weak Consider
A closer look…
A possible explanation for the Strong/Weak Consider results is that the chance of being admitted (based on admission index scores and LSP adjustments) should look like a parabolic curve (since it should be a normal distribution), and, based on these accuracy results, a curve with a high and narrow peak in the center (leptokurtic distribution).
Underrepresented Minorities (URMs)
Target admit rates for URMs, based on how LSP renders predictions: (same method as non-URM applicants after URM boost is applied)
When LSP said Admit: >= 87%
When LSP said Strong Consider: ~ 69%
When LSP said Consider: ~ 50%
When LSP said Weak Consider: ~ 31%
When LSP said Deny: <= 13%
Based on the 4,300+ LSN 2008-09 decisions for self-identified URMs and with LSP URM feature enabled:
When LSP said Admit, 963 instances: 87.1%, +0.1%
When LSP said Strong Consider, 418 instances: 70.8%, +1.8%
When LSP said Consider, 1034 instances: 55.3%, +5.3%
When LSP said Weak Consider, 436 instances: 33.2%, +2.3%
When LSP said Deny, 1468 instances: 14.9%, +1.9%
Difference between actual URM admit rate and predicted URM admit rate: (prediction categories weighted evenly)
Average rate: +2.3%
Median rate: +1.9%
Summary:
Prediction categories that were right on the money: Admit
Prediction categories that were reasonably close: Strong Consider, Weak Consider, Deny
Prediction categories that were somewhat off: Consider
@"Seeking Perfection" "The reason that the tool deals with URM status any better than other predictors is because the underlying data includes more finely detailed information on it, not some imaginary sophistication."
From the site:
URMs tend to have an improved chance of admission when compared to non-URM candidates with the same LSAT and GPA. URM races are generally considered to be African-American, Mexican-American, and Native American. There are arguments both for and against this practice, but the fact remains that it does exist. For the purposes of this prediction model, URM candidates receive a raw point boost to their index formula score; this boost is based off of a percentage of the roughly median applicant’s index score at a given school. This model gives the same percentage boost at all schools.
"I don't think I in anyway represented the search tool as not based on historical data or the law school predictor as based on something other than a fairly simplistic regression of that same data."
That ^^ doesn't sound that "imaginary" to me.
From the site:
Q: How did you create formulas for schools that don’t publish formulas?
A: A combination of regression analysis using 25%/75% LSAT and GPA data along with submitted user data and modifications to the resulting formula using submitted user data and law school applicant data available online. They’re not perfect at predicting, but then again, neither are the published schools’ formulas.
Q: Where do you find the data for all of this?
A: A number of information sources were used in compiling the data included in this model. These sources are: Internet Legal Research Group, Law School Admission Council, Law School Numbers, and US News & World Report. A number of Top-Law-Schools.com forum members have also kindly contributed information and feedback that has led to the ongoing improvement of this model. TLS forum member OperaSoprano has also provided valuable support and feedback; without OperaSoprano’s suggestion, I might have never created the part-time program prediction portion of the model. TLS forum member CyLaw has kindly offered a number of suggestions and support for Law School Predictor.
-So if they are using the same data, doesn't that contradict them using "more finally detailed information"- I think that would qualify as additional data? Also, can you explain what "finely detailed data" you were referring to?
From the site:
While the URM boost was not devised by using a particularly statistically rigorous technique (the URM feature debuted in Version 1.5, and has been tweaked a couple times since), it turns out that LSP is just about as good (or arguably better) at making predictions for URM applicants on LSN when compared to the average applicant on LSN. For those who sometimes suggest that decisions for URM applicants are largely unpredictable based on an applicant’s numbers, these results would shed some doubt on that assertion. This is not to say that URM applicant cycles are entirely predictable, but if a URM applicant plugs his/her numbers into LSP, s/he should get a decent of his/her chances.
The prediction categories used in testing accuracy are the same as LSP’s prediction engine with the exception of early prediction (which was not tested); predictions are adjusted to account for splitter-ness, weak GPAs, and URM status.
@"Seeking Perfection" Do you have any more baseless statements you would like me to disprove? Also, and again, would you please explain your qualifications/source of knowledge for all of your incorrect assertions and assumptions? Do you have a degree based in stats or numbers? Do you have any real experience working with large data sets? Have you ever modeled data in real world situations? I think I may already know the answer to those questions.. again that's my assumption
"LSN also has a tool for sorting through the data by LSAT score, GPA range, URM status, early decision, and which cycles of data you would like to include. It simply tells you how many people applied with your selected number range in that year, what percntage got in, what percentage of those that got in reported recieving scholarships, and what the average scholarship ammount was. I don't think that the law school predictor was doing anything all that much more impressive."
Or maybe you can explain how a regression model doesn't do anything different than that which you've described above? Is the percentage acceptance rate for variables the same thing as a probability of acceptance based on those variables?
Might be time to revisit the wikipedia page you got all your brilliant deductions from.
PS. I couldn't explain how irrelevant what you care and don't care about is to me. Please stop, I think you've had enough.
My credentials which I didn't mention because they are not relevant(beyond the facts that I know a bad regression when I see it, understand bias in a regression, and have the basic understanding of statistics you apparently lack) are that I am an Econemics Bachelor of Science with three published papers in Econometrics all dealing with moderately sized data sets in real world situations.
I don't have a problem with small data sets using excel (which I would hope is how the businesses you mention are using it), but it seems to me you just discovered why you shouldn't use it when you have serious large quantities data to deal with. My programming knowledge may be amateurish, but since you apparently slid through a less than quantitatively rigorous business program without picking up the knowledge that excel is not efficient at dealing with large data sets, it is still sadly beyond yours.
I wrote two responses to your ridiculous assertions that law school predictor had a regression analysis of uncommon sophistication and accuracy. You responded with 7 seperate responses which primarily attack me for not listing credentials which don't really matter here. It's not that I think your credentials are irrelevant. They objectively are irrelevant. Statistical facts are not changed by your how well or poorly credentialed your opinion is.
Finally, the reasons that I said the regression model isn't significantly better than the underlying data on LSN are a couple fold.
The first is that as you ought to know if your education in statistics was not woefully lacking, regressions are based on a standard set of assumptions. One of the most important ones is that there is zero covariance between independent variables and the error term. Due to extreme self-response bias that is definitely not true in the LSN data. There are ways to account for that, but they invariably increase the variance of your regression and in this case would have expanded the variance to the point where it would have been useless. Since they have a relatively narrow range to range to their predictions it follows that the law school predictor did not use a more robust estimator. This means that the predictions have a more narrow range, but may be wrong in some systematic fashion.
The reason I allowed that the two were nearly equivalent is that if you used the same range of data the predictions closely matched what you would get just from the LSN data. It therefore gives a ballpark for your chances of getting in which thankfully is all we need for application purposes.
I don't want to replicate the regression of the law school predictor or want 7sage to, not because I can't, but because the regression can't be a good regression as a result of the detailed, but biased data, because even if it was no one needs a perfect predictor of their law school chances to decide where to apply, and finally because it would take up time and resources which are valuable to me.
Other minor points that are less central:
I said they were using more finely detailed information than other predictors on the internet. I had the official lsat predictor in mind. You deliberately misinterpreted this as me meaning it was more finely detailed than LSN which is absurd since it's the same data.
It seems that after reading the website again you now agree with me that the law school predictor does not have a particularly impressive way of accounting for effects on URM's. You literally quote the website admitting this which may show you are capable of realizing you are wrong. They just used the LSN data which has that information on an individual level as opposed to other predictors not based on LSN which therefore don't have that sort of detailed data.
So in conclusion, I'm sorry that your education which apparently dabbled in statistics without explaining the requirements for a regression to work or teaching you how to run regressions on significant data sets. I'm glad that this has not handicapped your career too severely in business management. I wish you the best in law school.
And most of all I hope that you have convinced no one to waste any time, money, or energy making a new version of the law school predictor.
I wouldn't say I have had enough. But I can see a few reasons you might have said that you think I've had enough.
The first is that you thought that my failure to respond to your last seven messages was an indication I wouldn't respond and were either trying to bait me into responding or make it appear as though you had scared me away. I was going to respond anyway, but had other things to do so you need not bait me. And I think you will find that prospective lawyers are less likely than the general population to concede arguments to make you feel better about yourself. They need a good reason.
Thankfully, I have one. It is clear there is no interest in wasting 7sage resources on buying a program which can't provide accurate results, that I will not make one, and that you will need to learn how to use a more effective program than Excel before you make one. As such I'm not inclined to waste more time on this attempt at reeducating you or furthering your education in statistics.
If in your next seven posts, you manage to convince yourself that you were right, I made up the existance of statistical standards and hid them on a wikipedia page, and that the former law school predictor app was the greatest regression ever performed I'm okay with that. It's not ideal (that would be you reeducating yourself about statistics and learning some programming to better indulge your interest in regressions), but it is better than me wasting more time on it or 7sage wasting resources on it.
@"Seeking Perfection" Wow man, so you just stated you have never worked with data in the real world. Okay you understand how moderately sized data in a academic paper is not actually modeling data for real application. You talk a big game man, what are your papers? list them please? I don't believe they exist.. prove me wrong. Calling it now, you haven't even graduated have you?
Seems like your post is a bunch of different ways of trying to insult my education. I'm not sure if you don't understand why I mentioned that I model large data sets for a living for 6 years and have a degree in business stats not one that "dabbles in statistics".
You don't have to write a doctoral thesis every time you reply. Now your response on the other had had one actual fact regarding rudimentary assumptions of data sets for use in modeling, those assumptions have been accounted for in the data, if you actually took a second to look at the tool before rushing to insulting my credentials?
As to your thesis here:
P1- Insult/ admission you have no real work experience modeling data, and no experience working with large data sets.
P2- Insult/ assertions about proper modeling tools based off your 0 actual expereience
P3 insult
P4- Insult- my response indicated the variety of data, not just LSN data that was used in the regression model, also the accuracy rate I posted above completely disproves your assertion.
P5- WRONG- they use additional data sets and a index formula combo for.a regression. Again all this talk of stats and you can't seem to understand that? IDK why
P6- no one cares
P7-"The reason that the tool deals with URM status any better than other predictors is because the underlying data includes more finely detailed information on it, not some imaginary sophistication."- This is your own quote, no one is misrepresenting your ignorance. You did a good job of displaying that yourself. My response was to post the methodology employed in accounting for this factor which you deemed "imaginary".
P8- See responses indicating URM methodology.
P9- Insult in form of conclusion?
Line 11/12- Irrelevant
P13-Yes, I actually made several responses. Not to "bait" you but to systematically expose your ignorant assertions point by point. You just seemed to read the credentials bit?
P14 insult
P15- insult
So, how about you just go back and read my posts, each of which disputes and disproves every single statement you have made on my post in full. Now you can continue to sling insults or you can come with some actual facts? And maybe try and get a job while your at it? Seems like you can use a bunch or big words and take up 10 pages without offering a single actual basis for any of your assertions. As to mine, check the posts above, that's called proof.
@"Seeking Perfection" Also feel free to answer ANY, even a single one of the questions aside from your rudimentary display of intro data modeling assumptions that you learned in stats 101 or whatever. You're clearly delusional, please don't pollute my posts with any further insults, that seems to be 90% of your material, and avoiding any questions or aknowledgment of point by point destruction of your entire uneducated post, from beginning to end. Any other questions professor? Why do you not understand how me doing this for a living for 6 years, and having a degree in stats is relevant to discussing proper interpretation of statistical data models? That's like saying "just because you're degree is in math and you're a math teacher, doesn't mean you know math."
Bottom line, I posted the accuracy rate and the data sets used. How do you respond. Also, I think the fact that myLSN links to this predictor, in addition to a whole host of other law school prep sites, and the fact that it's licensed by top-law-schools.com speaks a bit to its relevance. Your clearly educated opinion notwithstanding. That enough posts for you? Want to offer an actual response now?
Hey guys....It's ok to argue but remember the number one rule on this forum:
1. Be nice.
https://7sage.com/discussion/#/discussion/15/forum-rules
@akistotle Bro read the OP no one is trying to argue, I have this clown on my post talking all kinds of baseless stuff and trying hard to insult me? I'm literally just going point by point smacking him down but he persists lol. You see this book he just wrote thats like 96% him coming up with different ways to try and insult my education while avoiding aknowledging that I disproved his every statement? I
Like @akistotle pointed out, the maxim of the forum is being nice. At this point, you guys are just kind of insulting each other and the original argument about the predictor has sort of faded into the background. It's never got to be about smacking anyone down or name calling. We all presumably share the same goal of mastering this test and gaining admission to a great law school.
@"Alex Divine" Look man, point being here is that this dude is making all kinds of statements that I am addressing point by point to be factually incorrect. I'm not about to sit here and let this dude post incorrect garbage all over my post. And I'm certainly not about to let this clown demean the tremendous amount of work that went into this model or talk trash about its accuracy when it has PROVEN rates. Talk trash about the data, without acknowledging its got MULTIPLE data sets. And lastly, talk trash about the value of a tool that has been corroborated by NUMEROUS law prep sites mentioned above and beyond those as well. Not to mention at this point this dude has insulted top 20 business programs, tier 1 consulting firms, THE FORTUNE 100 though?
I hear you, dude. Listen, calling @"Seeking Perfection" a clown or whatever isn't cool. He's a good dude and I've known him to be quite helpful on the forum. I also know you are a knowledgable and cool dude! So just chill with the name calling is all. I hear you... I really do. It's just not a big deal in the scheme of things. It's the internet and everyone is entitled to their opinion. You both make good points but the acrimony and ad hominems aren't getting us anywhere with this discussion; so let's just end it here.
If you want to chat more about maybe collaborating in the future on making our own predictor feel free to PM. We can talk about how the Cavs are doing so far this season as well
@"Alex Divine" No man it is a big deal when you directly insult people, this dude made a post that insulted me 15 separate times by my count. Directly demean the work of numerous people who are just trying to offer free resources without understanding the basics of different data displays and persisting on assertions that have literally been proven wrong multiple times. I'm sure he's a good dude, he need to learn to keep his mouth shut when warranted and acknowledge when every single one his points has been directly disproved.
@"Alex Divine" Bet money he doesn't have any published works.. liar.
@"Alex Divine" Can you highlight a single valid thing that he said in the 900 page long comedic insults he posted? Like literally, is there a single thing that he managed to prove here that wasn't directly refuted? Just wanted to see where your high degree of confidence is coming from with this "specialist" on this matter? This dude has never had a job, he doesn't even have a degree I'm pretty sure. Someone who insists that their own personal preferences about data views is statical law, who doesn't understand the relevance of a degree in statistics to discussing statistical data modeling? Who cannot tell the difference between a historical percentage and a probability? I called him a "clown" because every single "fact" that he posted makes me laugh, and he's so sure of himself too.
"One of the most important ones is that there is zero covariance between independent variables and the error term. Due to extreme self-response bias that is definitely not true in the LSN data."
Who seems to think that "self-response bias" makes this data unfit for regression modeling, but at the same time think this VERY SAME data is completely fine using the search tool on myLSN? Doesn't "self-response bias" mean people are lying? So then, how is the data on myLSN useful but not useful when modeled? And so by implication the data on lawschoolnumbers is incorrect, top-law-schools.com are idiots for owning the exclusive license to this tool. I literally posted the accuracy rates of this tool, and somehow he failed to notice or acknowledge how that accuracy could have accrued with such a flawed model?
Does any part of that seem logical or coherent to you? IDK maybe you could explain something that he's trying to say that I'm missing?
@"Seeking Perfection"
"One of the most important ones is that there is zero covariance between independent variables and the error term. Due to extreme self-response bias that is definitely not true in the LSN data. There are ways to account for that, but they invariably increase the variance of your regression and in this case would have expanded the variance to the point where it would have been useless. Since they have a relatively narrow range to range to their predictions it follows that the law school predictor did not use a more robust estimator. This means that the predictions have a more narrow range, but may be wrong in some systematic fashion.
"The reason I allowed that the two were nearly equivalent is that if you used the same range of data the predictions closely matched what you would get just from the LSN data. It therefore gives a ballpark for your chances of getting in which thankfully is all we need for application purposes."
MyLSN does not provide any "predictions". If the regression and search tool are "nearly equivalent", and you seem to think the regression is wildly inaccurate, how is it that you concluded that myLSN search is useful? Do you mind reviewing the post I made indicating accuracy rates and then explaining how it indicated the toll is "wrong in a systematic fashion"?
"Finally, the reasons that I said the regression model isn't significantly better"
"However, I think it is actually such a small improvement over the data searching tool already available on LSN that aquiring would be almost totally useless."
So you are simultaneously indicating that the data is biased because it is user reported, therefore a regression would be incorrect. However the same self-reported data which is bias by your own account is providing value through the myLSN search? How's that work? If that data is "bias" and that makes it inappropriate for modeling, wouldn't that also mean it would be inappropriate for determining anything at all? Or does is it not self-response bias in myLSN and then when its used for the regression, then it becomes self-response bias?
"But, when I say that I don't think it would be hard, that's on a relative scale. It would certainly take me either a significant amount of time or favors from friends."
So, would it be hard or not? You position is kind of unclear on that point.
"If in your next seven posts, you manage to convince yourself that you were right"
Or literally expose everything you just stated as incorrect? I just addressed every statement you made, every single one, please continue to hammer home your points. (Not actually sure what they are, you seem to be backpedaling and changing your responses)