$60 - tutor wanted

Link

Ninja say what!?!
#21
SEM final ehhh???? Mind telling me what you've covered in that course?

My professor bypassed what's currently known as the "classical" SEMs in favor of teaching us about the Non-parametric SEMs. Luckily, I had to use SEMs with another professor who was only familiar with the classical ones and got to learn about them that way. Kinda curious as to how widespread this is.
 

noetsi

No cake for spunky
#22
Since the offer is rescinded I figure it's ok to hijack this thread :) Noetsi we both know I'm a bit of a dope and I picked up R. There's different levels of understanding in R. To get the initial learning takes a couple of weeks to get down with a good guide (let me suggest Quick R and/or 'R in Action' a book by the same author.

After you get this stuff down you'll realize wow this does awesome stuff and you may move to other packages (I've went through ddply and reshape2 pretty thoroughly now and am working on tm and ggplot 2 currently).

When I think I think I've learned most of what it can do I see what bryangoodrich or dason or theEcologist can do and I realize there's even more it can do (so many packages so little time).

I see your intellect with stats so I really think R will be not too bad for you to undertake.

PS good luck with the SEM final
Thanks for all the kind comments, although I think you are confusing me with another poster :p but I still wish you were taking the SEM final not me. :)
 

noetsi

No cake for spunky
#24
SEM final ehhh???? Mind telling me what you've covered in that course?

My professor bypassed what's currently known as the "classical" SEMs in favor of teaching us about the Non-parametric SEMs. Luckily, I had to use SEMs with another professor who was only familiar with the classical ones and got to learn about them that way. Kinda curious as to how widespread this is.
This was the second course in a sequence specifically in SEM. But no non-parametrics (never even saw that mentioned in the SEM lit including books a lot more advanced than what I learned unless you mean the use of robust weighted least square as an estimator rather than ML or MLR).

This is what we covered:

A section on ML (the estimator most commonly used).
Non-normal and categorical data for SEM (how to deal with it).
Equivilent models
Full Structural Models
The use of mean structure
Multi-sample SEM (which uses mean structure heavily)
Between Group Differences in Latent Factor Means (basically multi-sample SEM with CFA not PA)
Latent growth models (time series with SEM effectively)
Non-linear latent growth models (plus a lot of speciality time series variations such as growth models with mediating variables or repeated measures)
Power

We used M+ which is supposedly the best SEM software available although it has some strange fetishes.
 

noetsi

No cake for spunky
#27
Well for example if you save your data in a different drive then the software is saved Mplus becomes dazed and confused and wonders aimlessly in the desert reporting strange errors that makes absolutely no sense. I lost a whole days work once that way. And you have to trick it to report certain useful information tied to the mean structure. Thankfully it is easy to trick (not surprising for something developed by a group of academics...).

My personal unfavorite is a policy of the company not the software. Generally graduate students can get software at a discount. Mplus has that as well, but if you have a doctorate already (even in a totally different degree area than the one you are working in now) you can't buy the cheaper software. I argued to no avail about that. Yet another reason to regret my phd. :mad
 

noetsi

No cake for spunky
#29
Yeah but all the profs like Mplus.....

Of course once IBM gets rid of R it won't matter much anyhow. I need to learn it quick before the hammer comes down.

Like most math and stats types Dason you are an idealist. You fail to consider the vast power of greed and fear (aka market forces).
 

spunky

Can't make spagetti
#31
Yeah but all the profs like Mplus.....
that is true indeed. the usual frontrunners when it comes to structural equation modeling software are LISREL (the prince among them all since it single-handedly saved structural equation modeling as a technique from dying at birth), AMOS for SPSS, EQS and Mplus now, with that said. the latest gossip:

AMOS: makes pretty drawings ready for publication but it's extremely cumbersome and very restrctive.... plus i cant find the link but i know a Belgian group developed a LaTeX thing to draw structural equation models (making AMOS mostly irrelevant now). if anyone out there can find where the page is please post the link because i lost it :(

EQS: peter bentler (main developer) is retiring so it'll go the way of the dinosaur...

LISREL/PRELIS: had its moment but no longer the dominating force

Mplus: the shining star... (for now)

problem with Mplus: great for data anlaysis, horrible for everything else. nevertheless, Muthen (the creator) who is a statistician himself knows the power of R all too well and it's very easy to interface Mplus through R. anyways, if you really wanna exploit the power of structural equation modeling in R you'd use OpenMX. MX used to be another commercial software for structural equation modeling but in what i think an act of great kindness, instead of closing shop and disappearing once they stopped being popular, the developers wrapped everything up in an R library and it is now the most powerful R library to run SEM models. pair that up with all the other extra packages for R and Mplus ends up being the useless pile of goo...

@ noetsi: a great book for logistic regression would be Applied Logistic Regression by Hosmer & Lemeshow. that's my default reference book for everything logistic and the emphasis is on applied ;)
 

Link

Ninja say what!?!
#33
Of course once IBM gets rid of R it won't matter much anyhow. I need to learn it quick before the hammer comes down.

Like most math and stats types Dason you are an idealist. You fail to consider the vast power of greed and fear (aka market forces).
Wow. Really??? So by the same argument, Linux will disappear too???
 

Dason

Ambassador to the humans
#34
I wouldn't be too worried. noetsi likes to troll this idea on us every now and then.

I don't really think they see how incredibly impossible it would be for a company to buy out R and then drive it into the ground. As much as noetsi believes in the power of greed and fear - we will always have GNU fighting for free software. And since R is protected by a copy-left type license with the license holders being adamantly for open source software... I don't really see any company buying out R (IBM especially). But I've just come to ignore these kinds of comments for being absurd. noetsi can call me an idealist (and I'd like to think I am) but I also truly think that the only thing that could happen to R is that a newer open source statistical package could become popular and overtake R - essentially killing R in the process
 

trinker

ggplot2orBust
#35
R and linux do play on greed: I want the best software and I don't want to pay for it(different kind of greed). Plus I think people do it just in opposition of corporations, "look we can do what you do better and it's free".

I just started using linux ubutnu. If mac users knew what this was and could do I bet they wouldn't drop the money they do on a new mac machine (unless they just buy it because it's trendy; which I suspect may often be the case).
 

Dason

Ambassador to the humans
#36
R and linux do play on greed: I want the best software and I don't want to pay for it(different kind of greed). Plus I think people do it just in opposition of corporations, "look we can do what you do better and it's free".
Some of it is greed. Some of it is trying to create a better world. {Edit: I'll add that I don't mind paying for software - people put a lot of hard work into some products and giving them money seems perfectly reasonably to me. But I like having the freedom of changing the code if it doesn't do what I want it to do}

I just started using linux ubutnu. If mac users knew what this was and could do I bet they wouldn't drop the money they do on a new mac machine (unless they just buy it because it's trendy; which I suspect may often be the case).
I bet a few mac users would switch but there are more reasons to use a mac other than that they're trendy. One reason is the software. There is quite a bit of mac only software that is great for certain things - mainly the artistic types of things; the system is highly integrated as well so you get a very cohesive experience from operating a Mac. And while there are Linux software 'equivalents' to the mac/windows only software a lot of times they don't do as good of a job. I'm a realist in this regard.

One other thing that is nice about macs is you actually get some support from a company. Go to an apple store and tell them your computer is not working - you'll probably have it fixed relatively soon and you didn't need to do any messing around with it yourself. If your computer with linux suddenly starts giving you a kernel panic every time you boot up ... who are you going to call or take it into to get help? Sometimes it's not too bad to work out yourself but if you're absolutely terrified of the command line then good luck getting anything done.

Linux is great but I don't see it overtaking the market in the next five years. Eventually? Maybe... but we would need a distro that is completely user friendly and does the tasks that most users want to do even better than what they're using right now (windows or OSX). Because if there isn't an incentive to switch (and being free isn't enough of an incentive to switch when you're essentially throwing out the OS that comes with the computer) then it's hard to get people to try something new.
 

trinker

ggplot2orBust
#37
Dason said:
There is quite a bit of mac only software that is great for certain things - mainly the artistic types of things;
Valid points (my firend Joe using final cut and all sorts of mac programs for videos and music) but many(I'd venture to say most) of the people I see from the university carry them around and aren't artsy people. They then use macs for for microsoft purposes (ie office, endnote, and nvivo). Then these programs (particularly nvivo) don't run properly or at all on the mac and then they partition the hard drive and run windows anyway. They didn't buy em cause they're artsy and they didn't buy em for pragmatic reasons (hence the need to partition), why then did they buy em? Trend.

They then go on to blame mac or the program for not running appropriately when they should have been investigative in their purchase in the first place and bought based on intended purpose, then they'd wind up more satisfied with their product and probably 1500 bucks richer (aver $ of a pc laptop - aver $ of a macbook). Plus they can use the savings to upgrade every few years when the pc goes obsolete or the mother board is fried :)