Research Question

[Therapist4Chnge]

---

Members do not see this ad.

The only thing I can think of is, "If A = B, and B=C, then A=C"....something like that.

 

[JackD]

Yes again. You're becoming quite the stats guru.

The stats is strong with this one.

Apparently so. I got the results of the exam a few minutes ago and it was one of the highest scores i ever received on a psychology test. One of the biggest shocks of my week (if you ignore the democratic primaries from yesterday).

 

[biogirl215]

Okay, a question for all you design experts:
How would you construct an experimental/causational study on whether technology-based relationships negatively affected mental health?

I know someone who wants to do this experiment (or would, if they had any free time at all, which they currently don't), and I honestly can't wrap my head around how you could measure that causationally. Any ideas?

 

[empathiosis]

Okay, a question for all you design experts:
How would you construct an experimental/causational study on whether technology-based relationships negatively affected mental health?

I know someone who wants to do this experiment (or would, if they had any free time at all, which they currently don't), and I honestly can't wrap my head around how you could measure that causationally. Any ideas?

I'll take a stab at part of it (even though I should be doing work). First you would have to define "technology based relationship". Do you mean people who have relationships with people they have never met f2f and only know through IM or email or games? Second, you would have to define what you mean by negativelly affected mental health ... and then if you only got correlational results, in other words, people who have more online relationships tend to be depressed, you would still have to worry about the underlying variable, which might be people who are shy or who lack social skills lack social support therefore are more likely to suffer from depression. Get my drift?

There has been a lot of work done on internet addiction. I also know in the early days of the web, there was a classic study which seemed to show that length of time spent on the internet was related to depression. Here's the citation for that:

Kraut, Patterson, Lundmark, et al. Internet Paradox: A social technology that reduces social involvement and psychological well-being?American Psychologist, Vol. 53,


But when Kraut et al. went back and studied the original families used in their study they found normal levels of depression and attributed the initial effect to their overuse of something which was novel.

It's a tough experiment to do elegantly.

 

[JockNerd]

Okay, a question for all you design experts:
How would you construct an experimental/causational study on whether technology-based relationships negatively affected mental health?

I know someone who wants to do this experiment (or would, if they had any free time at all, which they currently don't), and I honestly can't wrap my head around how you could measure that causationally. Any ideas?

I can only imagine that being done longitudinally by setting up some sort of fake long-term online relationship. I don't know how interesting that would be. I also think it would be impossible to get irb approval to do something causal if you think the result of your manipulation is long term degradation of mental health.

 

[empathiosis]

sorry, my brain isn't working right today. although you clearly wrote experiment/causation, for some reason i thought correlational was an option. even if an irb did allow you to set up pairs to have online relationships ... there are too many other confounding variables. unless of course you controlled for prior experience with mood disorders, etc. by selecting participants with low scores on something like the BDI.

JockNerd is correct though in most ways.

I doubt though that this experiment would lead to degradation of mental health because I don't think that having online relationships causes depression. It would be a lot easier to do what Kraut et al did, which was to introduce the internet to families that did not previously have access and to measure mood after the period of exposure. Have your friend take a look at that article. It may give him/her more ideas.

 

[JackD]

I'm am somewhat lost on something for the research project we are doing for research methods course.

I am trying to describe a 2x2 analysis of variance interaction. The level of career certainty is the dependent variable and there is an identity status/sex interaction. So here is how this looks

Career Certainty

Pure Diffusion

Males (m=14.40)
Females (m=15.188)

Pure Moratorium

Males (m=14.389)
Females (m=19.541)

I am wondering how to describe this. Do i compare the differences between the two means? Should i say something like

"there is a .8 difference in means between males and females who are categorized as diffusion and a 5.2 mean difference between males and females categorized as moratorium. This shows that identity status does have an effect on career certainty for females, while for males, it makes no difference."

Is that generally what i want to say or am i missing the point of these interactions?

 

[JockNerd]

I'm am somewhat lost on something for the research project we are doing for research methods course...

Without standard deviations and, ideally, ranges, you can't really say anything. If, for example, if whatever those means represent are on scales that range from -100 to 100, and each group has a standard deviation of 45, then the differences are inconsequential. If, on the other hand, the scales range from, say, 10 to 20, and the standard deviations are .2, then those differences are enormous.

If it's just a poor question and not a trick (or you were given but just didn't post the SDs), then I think your answer is correct.

 

[JackD]

Without standard deviations and, ideally, ranges, you can't really say anything. If, for example, if whatever those means represent are on scales that range from -100 to 100, and each group has a standard deviation of 45, then the differences are inconsequential. If, on the other hand, the scales range from, say, 10 to 20, and the standard deviations are .2, then those differences are enormous.

If it's just a poor question and not a trick (or you were given but just didn't post the SDs), then I think your answer is correct.

I can tell you what the standard deviation is but i have no clue what the range is.

Pure Diffusion

Males (m=14.40) (sd=.844)
Females (m=15.188) (sd=.600)

Pure Moratorium

Males (m=14.389) (sd=1.348)
Females (m=19.541) (sd=.684)

 

[JockNerd]

looks to me like you're doing it correctly then.

 

[JackD]

looks to me like you're doing it correctly then.

Thank you sir. That is actually a great help.

 

[JackD]

I'm trying to study for the final in my research methods class but there is one subject that is pretty confusing.

I am reading a chapter about 2x2 factorial experiment. When it is talking about the "overall main effect" do they mean the average of the two main effects for an independent variable? And when they talk about "interactions" do they mean the difference between the two main effects in the same independent variable or is it the difference between the same manipulation across the two independent variables or is it when you look at the difference between all four main effects?

 

[JockNerd]

I'm trying to study for the final in my research methods class but there is one subject that is pretty confusing.

I am reading a chapter about 2x2 factorial experiment. When it is talking about the "overall main effect" do they mean the average of the two main effects for an independent variable? And when they talk about "interactions" do they mean the difference between the two main effects in the same independent variable or is it the difference between the same manipulation across the two independent variables or is it when you look at the difference between all four main effects?

Here's a good site with systematic explanations of the differences between single main effects, two main effects, main effects and interactions, etc. Should answer your question about interactions. http://www.psychstat.missouristate.edu/multibook/mlt09m.html

Does the textbook use the words "overall main effect"? I don't ever recall hearing that before. Each IV would have a significant or nonsignificant main effect. Maybe it means the model intercept? But I think that's almost always significant and not super important...

 

[JackD]

Here's a good site with systematic explanations of the differences between single main effects, two main effects, main effects and interactions, etc. Should answer your question about interactions. http://www.psychstat.missouristate.edu/multibook/mlt09m.html

Does the textbook use the words "overall main effect"? I don't ever recall hearing that before. Each IV would have a significant or nonsignificant main effect. Maybe it means the model intercept? But I think that's almost always significant and not super important...

The problem is that the book gives a very poor definition of what a 2x2 interaction is, not explaining what it means, just by giving an example of what one is. They just give four different experimental conditions. I don't know what the independent variables are, i don't know how all of the conditions are even related. It is like trying to teach someone multiplication by just rambling off a bunch of examples of multiplication problems.

They just give a convoluted example and then they move on to explain the features of a 2x2 without ever really explaining what a 2x2 interaction is in the first place. I spent the last hour reading the same three pages over and over trying to decipher their definition but i have no clue what they are trying to say. The whole chapter is a mess. Hopefully your link will be better at explaining all of this. I will have to look it over when i am not so frustrated.

 

[JackD]

Ok, i am still a little lost but i think i can figure it out through my crappy class notes. I just need to know if this is correct

In the class notes, the professor gave an example of a 2x2 factorial experiment. It compared empathic therapy for men and women with behavioral therapy. In an experiment like that, empathic therapy is one independent variable and behavior is the other, correct? Then the two levels are men and women?

The main effects then would be taking the average of how men and women scored for empathic and the average of how they scored for behavior?

Please tell me that is right because it actually makes sense to me. If it isn't, i'll probably end up sitting in the corner, rocking myself back and forth.

 

[Ollie123]

I'm not clear on what you are saying - it doesn't mean you're wrong, I just might not be understanding correctly. I'll say how I would interpret that experiment, and you can compare that with what you meant to say😉

The premise of a 2x2 factorial is you have 2 different IVs.
IV 1 = Type of therapy (Levels: Empathic versus behavior therapy)
IV 2 = Gender (Levels: Male versus female)

A main effect of therapy would be if behavior therapy worked better (overall, for everyone) then empathic therapy.

A main effect of gender would be if women improved more than men regardless of therapy.

An interaction would be if women did better with empathic therapy, and men did better with behavior therapy. Actually it could be a number of different things, but the key is that the effects of one IV are different for varying levels of the other IV.

Is that what you meant? I thought you were thinking along the right lines, but like I said, I couldn't tell for sure.

 

[JackD]

What i was thinking was

IV 1=Empathic with men and women as the levels
IV 2=Behavioral with men and women as the levels


What i have in my notes is a table like this.

 

[Ollie123]

Ahh, close but not quite.

"Levels" exist within each variable. Its a little confusing at first, but you get the hang of it.

The variable is the broader concept (e.g. Type of treatment). The level is just the specific value of that variable.

 

[JackD]

Ahh, close but not quite.

"Levels" exist within each variable. Its a little confusing at first, but you get the hang of it.

The variable is the broader concept (e.g. Type of treatment). The level is just the specific value of that variable.

I am so glad i'm not trying to get into research.

I think i get it, just one more question. In the table i posted, there could also be main effects for each gender correct? You could compare men in one group to men in the other and the same for women.

 

[Ollie123]

Nope. Well, I mean you COULD test that, but then you're no longer operating within a 2x2 design.

Main effects - only considering 1 IV
Interactions - Multiple IVs

That's vastly oversimplifying things, but I think it should suffice for most undergrad-level classwork.

 

[JackD]

Here were, about ten months later, and I still don't have this totally down. However, this is a fairly simple question, so I don't feel too down.

Right now I am putting the final touches on my first research paper 🙂sleep🙂 and there is just one thing I am not clear on. The professor lectured on the basics of writing the paper in APA format and in my course notes, I wrote down that in the results section, for all of the statistical information you provide, the independent and dependent variable need to be specified. For example, if you have three correlations that need to be presented, you have to say "For the first correlation the independent variable was...and the dependent variable was....For the second correlation, the independent variable was.....and the dependent variable was......For the third correlation, the the independent variable was......and the dependent variable was....."

Is that the way it has to be done or are my notes incorrect and is that unnecessary?

 

[JockNerd]

Here were, about ten months later, and I still don't have this totally down. However, this is a fairly simple question, so I don't feel too down.

Right now I am putting the final touches on my first research paper 🙂sleep🙂 and there is just one thing I am not clear on. The professor lectured on the basics of writing the paper in APA format and in my course notes, I wrote down that in the results section, for all of the statistical information you provide, the independent and dependent variable need to be specified. For example, if you have three correlations that need to be presented, you have to say "For the first correlation the independent variable was...and the dependent variable was....For the second correlation, the independent variable was.....and the dependent variable was......For the third correlation, the the independent variable was......and the dependent variable was....."

Is that the way it has to be done or are my notes incorrect and is that unnecessary?

It doesn't have to be done that way. But if the prof wants you to do it, it might be more as a demonstration that you know which is which, so if the prof said to do it I'd just do it, in case it's on the grading checklist.

 

[JackD]

It doesn't have to be done that way. But if the prof wants you to do it, it might be more as a demonstration that you know which is which, so if the prof said to do it I'd just do it, in case it's on the grading checklist.

Good point. It will be a pain but I should probably play it safe. Thanks.

 

[Ollie123]

I agree with JN you should do it that way just for the sake of grading.

In practice, no one actually writes it exactly like that. Its usually pretty obvious what the IV and DV are, but it IS important to make it abundantly clear what you are testing. For the most part I think people do a good job with this, but I've occasionally caught vague statements in results sections where you can't be 100% sure what they are referring to and its a huge pain to try and figure it out.