Fleeting/Random Thoughts

spunky

Doesn't actually exist
But isn't planning a usable sample size part of sitting down and planning a well-designed experiment? I think it's emphasized too much as the only thing that matters because there are definitely aspects of a study designed that don't get considered nearly as much as they should but sample sizes still plays an important part don't you think?
yeah... i guess i didnt explain myself correctly there. you certainly have a point in the fact that you need to have some idea of how many people, dogs, galaxies, neurons or whatever you're gonna measure but, as you said, at least here in the world of social/behavioural/health sciences, people line outside my office asking for power analysis this and power analysis that and if they're "guaranteed" to get significance if they can get their sample size right and i'm like "no, no, no... focus on your research, on what you should be doing... the research question should guide the analysis and not vice-versa" now, i understand where they're coming from in a way because nowadays a lot of people ask for power stuff to even consider an application for funding, but still... it's weird here how people sometimes grab on to something for dear life and change it into something it was not meant for (like factor analysis, people pretty much get married to their factors)... one aspect that i would love people pay more attention to are confounders or missing variables instead of only sample size and statistical significance... i've been doing a lot of research on suppression effects in multiple regression and i think it's surprising, at least to me, the lengths people go to dump or just blatantly ignore suppression effects in the social sciences...

how's the RA position going, btw?

Dason

The RA position is interesting. There are some unique challenges involved with the data and I might end up seeking help from a couple of users here in the future. It should be fun.

spunky

Doesn't actually exist
hey, glad to hear that... at the risk of sounding too nosey could i ask what are you doing? or what kind of data analysis are you doing?

Dason

Right now I'm working with some RNA-seq data trying to find differential expression between control snakes and heat stressed snakes . The design of the experiment (which we weren't consulted on but logistically they were constrained to this design) is less than optimal (story of all research right?).

I'm working solely in R though which makes me happy. When I was consulting I worked almost solely with SAS which is not my platform of choice.

trinker

ggplot2orBust
2. Why are peer reviewers in psych more interested in pushing their pet theories than, I don't know, actually making useful comments about data analysis?
Peer reviewers actually read the article that's sent to them? You've got one up on others then even if they are pusing pet theories.

CB

Super Moderator
Peer reviewers actually read the article that's sent to them? You've got one up on others then even if they are pusing pet theories.
Lol! Guess I should be thankful for small mercies

TheEcologist

Global Moderator
Wow! Congratulations Dason! And how did it go Link?

I am getting to the point that I should pop the question soon too.... in the next few months I should try and find the courage

Tricky questions like when, where and how are still bothering me though....

TheEcologist

Global Moderator
Good news: Just got a new house! Got the keys on Friday! Finally moving from a one room apartment to something with a garden..
Bad news: Just when I started working on RBattleship, now I will be spending my free time painting and building stuff

Dason

Well that sounds awesome. Congratulations! I can't wait to get a house but it'll probably be quite a few years until that happens.

By the way: a few friends of mine saw my minesweeper clone and loved the idea of implementing games in R. One wants to make a version of "Guess Who" but I'm not sure if they'll actually follow through with that. I also don't know exactly what they planned on doing for the AI of the opponent. I mentioned that to them and said that it really only works well as a multiplayer game without some amazing AI. Then he went on to say "well why not just make it multiplayer". I don't think he really thought much about it...

I worked on the gWidgets version of minesweeper and implemented an option to right click to place flags. I also finally figured out how to change the color of glabels so the numbers can be colored. Sometimes I wish gWidgets was documented a lot better than it currently is (and it's not even that bad of documentation)... there's just a lot there that they don't mention at all in the documentation or the vignette.

Lazar

Phineas Packard
Lol! Guess I should be thankful for small mercies
bah I just got a review back from a guy who said more or less "how come you did not use this theory. I know it is not based on the area you are studying but....." Drives me bonkers. Trying to figure out how to tell the guy "I did not use the theory because it is irrelavent to this area of research" but in a nice way.

Dason

Well just because it's irrelevant doesn't mean you can't use it. Just to be safe you should also stick in some SEM and some HLMs and some GLMs. Then you'll have your bases covered.

Lazar

Phineas Packard
lol actually it is a sequential logit model using latent predictors with clustered data so I already have all those bases covered

CB

Super Moderator
here in the world of social/behavioural/health sciences, people line outside my office asking for power analysis this and power analysis that and if they're "guaranteed" to get significance if they can get their sample size right
This attitude to research really bugs me. Not only is the enterprise oriented towards confirming theory rather than subjecting it to potential falsification, but the researcher actually wants a guarantee that they'll get the result that they want?

I wonder if it's worth asking: But what if your theory is wrong? How will we know that it's wrong? And if it is wrong, does this study you're planning have any realistic chance of showing that your theory is wrong? (remembering that a non-significant p value isn't falsifying evidence, just inconclusive)

IMO figuring out how to subject a social science -type theory to genuinely risky tests with the potential for falsification is rather harder than power level calculations.

(Ok, so I've been reading a bit of Popper lately... )

TheEcologist

Global Moderator
By the way: a few friends of mine saw my minesweeper clone and loved the idea of implementing games in R. One wants to make a version of "Guess Who" but I'm not sure if they'll actually follow through with that. I also don't know exactly what they planned on doing for the AI of the opponent. I mentioned that to them and said that it really only works well as a multiplayer game without some amazing AI. Then he went on to say "well why not just make it multiplayer". I don't think he really thought much about it...
Oke some random thoughts: The 'multiplayer' option may be more interesting than I first realized. You know it may be useful if there was a way to run two R sessions on multiple pc's simultaneously in one synced workspace (besides it use for gaming), you establish a connection between two machines and you can work on data with your collaborators/supervisor simultaneously. This kind of thing is getting pretty popular and can be darn useful when your collaborator lives in another continent. Guess it would work through VPN.. or something.

Dason

Nice! My wife and I were looking for ways to get some additional income. I've been wanting to do some more teaching for a while (I love my RA but teaching is something I miss terribly). What do I hear about today? An elementary statistics course that needs a teacher at the hospital! I talked to the contact here at the department and I got a number to call and it looks like I got the job. I guess the person they was going to teach the course had to cancel at the last minute so they were pretty desperately looking for somebody.

*Excited*

TheEcologist

Global Moderator
Nice! My wife and I were looking for ways to get some additional income. I've been wanting to do some more teaching for a while (I love my RA but teaching is something I miss terribly). What do I hear about today? An elementary statistics course that needs a teacher at the hospital! I talked to the contact here at the department and I got a number to call and it looks like I got the job. I guess the person they was going to teach the course had to cancel at the last minute so they were pretty desperately looking for somebody.

*Excited*
Congratulations! I bet you will be a great teacher, through teaching the basics I have learned so much. its surprising how much you can learn from questions of people who are trying to understand stats

Today, I am just frustrated that those grant deadlines always seem to sneak up on you ;-) Its evening and I'm still at the university.... oh well.

Dragan

Super Moderator
Nice! My wife and I were looking for ways to get some additional income. I've been wanting to do some more teaching for a while (I love my RA but teaching is something I miss terribly). What do I hear about today? An elementary statistics course that needs a teacher at the hospital! I talked to the contact here at the department and I got a number to call and it looks like I got the job. I guess the person they was going to teach the course had to cancel at the last minute so they were pretty desperately looking for somebody.

*Excited*

Teaching Statistics in a hospital???...Now, there's one I haven't heard of

Who are the students?

trinker

ggplot2orBust
I'm trying to write an [R] code to bake me some cookies. I just can't get it to work. Here's what I have so far; any suggestions?

Code:
bake.cookies<-(type="choc.chip"){
switch(type,
choc.chip=,
oatmeal.raisin=,
peanut.butter=,
sugar=,)
}

Dason

If you had an automated cookie maker connected to your computer that you could control through the command line you could use the "system" command to control it.

trinker

ggplot2orBust
Dason,

bake.cookies<-function(type="choc.chip"){
}