Multiple Regression - reporting results

Ask here about academic and research issues, like designing studies, recruiting participants, choosing statistics, submitting for publication, etc.

Multiple Regression - reporting results

Postby Frozengiblets » Fri Jan 09, 2009 2:10 pm

Hi,

Is there anyone out there who can help clarify what exactly should be reported in the results section of a report after running a multiple regression test?

I think I need to report these values:

• adjusted r square value
• standard error
• F value
• Significance
• Beta coefficient

But whether they should be within a table or not.. and whether a graph or descriptives are required, I haven't a clue!

Could you offer me some tips on what exactly to report? I'd be extremely grateful.

Many Thanks
Frozengiblets
 
Posts: 5
Joined: Thu Dec 18, 2008 7:30 pm

Want to advertise with us? You or your business could fill this space. Email clinpsyforum@gmail.com for information.

Postby Bubbly » Fri Jan 09, 2009 4:01 pm

I reported the multiple regression like this in my undergrad dissertation:



In order to discover what might explain the significant overlap between the dissociation and schizotypy measures, multiple regression analysis was applied. In two separate analyses using the enter method , non-pathological dissociation (DES) and modified schizotypy (O-LIFE) scores were entered as dependent variables with the everyday memory failures (EMQ) and sleep-related experiences (ISES) scores entered as a block of predictor variables. They were entered this way due to hypothesis 5.

3.5.1. The relationship between dissociation and the independent variables
Image

Table 5. reveals a significant model for the predictor variables with a multiple correlation of .52, [F(2,119) = 21.720, P < .001; adjusted R² = .258]. However, Table 5. also shows that the ISES scores were not a significant predictor, but EMQ scores were. This indicates that the everyday memory failures scores were important for a significantly better prediction of non-pathological dissociative tendencies.



Figure 1.
A scattergram showing the relationship between schizotypy (O-LIFE) and dissociation (DES) measure scores and their significant predictor, everyday memory failures (EMQ) measure scores
Image
User avatar
Bubbly
 
Posts: 323
Joined: Sun May 20, 2007 8:01 pm
Location: England

Postby Ruthie » Fri Jan 09, 2009 5:11 pm

Can't add much to what Bubbly has said, but are you understanding what the values mean and how to interpret them? I just know that when people are struggling with results, they're not sure what everything means.

Ruthie
User avatar
Ruthie
Site Admin
 
Posts: 2968
Joined: Sat Mar 24, 2007 11:32 pm
Location: London

Postby Frozengiblets » Fri Jan 09, 2009 5:42 pm

Hi thank you for your inputs, Yes I do understand what the values mean - I'm just not confident on how to display results in an appropriate format.

My model was not significant and only 1 variable was found to be a significant predictor (eek!). Is it worth displaying a graph?

Also :
- do I have to report a predictive regression equation? I'm not sure if this is considered superfluous.
- are there any descriptives I need to report e.g. mean, SD?
Frozengiblets
 
Posts: 5
Joined: Thu Dec 18, 2008 7:30 pm

Postby Ruthie » Fri Jan 09, 2009 6:22 pm

Frozengiblets wrote:My model was not significant and only 1 variable was found to be a significant predictor (eek!). Is it worth displaying a graph?


I would do a scatterplot displaying the correlation between that variable and the outcome variable.

Also :
- do I have to report a predictive regression equation? I'm not sure if this is considered superfluous.


Most people don't, especially in psychology as any regression will only account for a small amount of variance so prediction equations aren't all that helpful.

- are there any descriptives I need to report e.g. mean, SD?


Always report descriptives for any stats.

Ruthie
User avatar
Ruthie
Site Admin
 
Posts: 2968
Joined: Sat Mar 24, 2007 11:32 pm
Location: London

3d scattergrams/ multiple regression

Postby moonbeam » Wed Feb 09, 2011 12:25 pm

Could anybody help me with my multiple regression at all? Any help appreciated.

I have selected cases and got a random sample of 100 cases (from 300). "research has indicated that intention is the strongest predictor of behaviour followed by perceived behavioural control"

DV - Behaviour
IV - Intention
IV - Perceive behavioural control

I need to run an appropriate statistical analysis to determine the predictive value of both variables on the data.
I also need to predict a score from it too, i have the equation for this i just need to run the statistical test to find out the numbers.

In my opinion (which may be wrong as this is my frist try at multiple regression) is to run a linear regression using my DV and two IV's then to do a scattergram?
Due to using both IV's I thought the 3d scattergram Bubbly used was the most appropriate? However my values appear wrong? and I can't work out how to input the plane of best fit?

If anyone could shed some light on this for me it would make me very very happy! Reading too many stats books has left me baffled and confused :(
There may not always be a solution to the problem, but there is always another perspective...
User avatar
moonbeam
 
Posts: 552
Joined: Fri Apr 09, 2010 11:37 pm

Postby LaLeonessa » Thu Feb 10, 2011 3:28 pm

I won't comment about the appropriate kind of chart because I really don't know (I have just been using line graphs), but you need to be a bit more specific about your question as "the predictive value of both variables" can be taken a few ways. You can...

(1) Look at Control's effect on Behaviour, and Intention's effect on Behaviour, separately. This is a simple linear regression of IV=Control DV = Behaviour, rinse and repeat for your other IV.
(2) Look at Control's effect on Behaviour controlling for Intention, and Intention's effect on Behaviour controlling for Control. For this you enter in Step 1 both IVs. The output will show you e.g. Control's effect on Behaviour once Intention has been accounted for.
(3) Look at your IVs in interaction. First create an interaction variable using SPSS of Control X Intention. In Step 1, enter your IVs separately as in (2) but in Step 2 enter your interaction variable. The output shows you the main effects of your IVs controlling for the other variables, as well as the interaction effect. Usually we are very interested in interactions.

This is it at its most simplistic (though it's definitely not simple! I've so been there, with the battle scars to show!).
LaLeonessa
 
Posts: 122
Joined: Mon Apr 21, 2008 12:49 pm


Return to Academic/Research

Who is online

Users browsing this forum: No registered users and 1 guest