The Student Room Group

Psychology stats help!- I've been told it's a regression but I don't see why? :(

Hi, so for my research project, my supervisor has suggested a multiple regression analysis, but I don't really understand why it is a regression, I feel like it's an ANOVA. So can someone please offer some advice here, I hope I'm wrong and it is in fact a regression!

This is a quick outline of my study. It's about the effect of information framing on attitudes. So it's a pre and post test design, I measure attitudes at time 1, I present 3 different types of information to 3 different groups of participants, and then the participants repeat the attitude scales at time 2. So my hypothesis is that presenting one type of information will lead to more attitude change, than the other two.

However, what's further complicating this, is that I'm also looking at an interaction, where I hypothesize that people's change in attitudes may depend on their level of individualism. Now this interaction is really confusing the issue for me. Because I'm looking at a within-subjects pre-post design, which is also compared against the three conditions, and I'm also looking into whether the change in attitudes depends on the level of individualism.

Now my research supervisor knows what he is doing, and his group has conducted something similar recently, but it just doesn't feel right to me. Can anyone help me out please, and offer some advice as to why it is a regression? :frown:
Here's what I think, generally I am wrong about this things but like to have a go and then wait for @iammichealjackson to come and correct me.

You have two hypotheses
a) Does the stimuli affect attitudes - this requires an ANOVA to analyse
b) Does the level of individualism affect attitudes (how are you measuring individualism by the way?) - this requires a regression to analyse
Reply 2
Original post by _Sinnie_
Here's what I think, generally I am wrong about this things but like to have a go and then wait for @iammichealjackson to come and correct me.

You have two hypotheses
a) Does the stimuli affect attitudes - this requires an ANOVA to analyse
b) Does the level of individualism affect attitudes (how are you measuring individualism by the way?) - this requires a regression to analyse


Thanks a lot for your reply, it is much appreciated!

I agree that a) is an ANOVA, and if it wasn't for the interaction I would be happy with that. But my supervisor seemed to think that both hypotheses could be tested using a regression, but I don't really see how that would work.

Also, I'm just reading into the assumptions of a multiple regression right now, and apparently it doesn't work if you have Likert scales as a dependent variable, and that's how I've measured attitudes? So apparently I'd have to conduct an ordinal regression, which to be honest, up until now I didn't think existed!

I've also measured individualism using a likert scale.
Reply 3
Can anybody else possible help me out please?
Original post by Twinpeaks
Hi, so for my research project, my supervisor has suggested a multiple regression analysis, but I don't really understand why it is a regression, I feel like it's an ANOVA. So can someone please offer some advice here, I hope I'm wrong and it is in fact a regression!

This is a quick outline of my study. It's about the effect of information framing on attitudes. So it's a pre and post test design, I measure attitudes at time 1, I present 3 different types of information to 3 different groups of participants, and then the participants repeat the attitude scales at time 2. So my hypothesis is that presenting one type of information will lead to more attitude change, than the other two.

However, what's further complicating this, is that I'm also looking at an interaction, where I hypothesize that people's change in attitudes may depend on their level of individualism. Now this interaction is really confusing the issue for me. Because I'm looking at a within-subjects pre-post design, which is also compared against the three conditions, and I'm also looking into whether the change in attitudes depends on the level of individualism.

Now my research supervisor knows what he is doing, and his group has conducted something similar recently, but it just doesn't feel right to me. Can anyone help me out please, and offer some advice as to why it is a regression? :frown:


I'd be happy to help another twin peaks fan! Firstly, ANOVA and multiple regression are essentially the same test some people just argue that ANOVA is just a special case of regression (e.g. http://www.theanalysisfactor.com/why-anova-and-linear-regression-are-the-same-analysis/). I don't think you can have continuous predictor variables in ANOVA, which makes it limited compared to regression.

Secondly, likert scales work equally well for ANOVA and regression. If you have a decent (>20-40 sample size) and a decent number of levels on the likert scale (>3-4) then central limit theorem says that you can assume that the sampling distribution outcome variable will be normal (which is what matters).

Thirdly, this sounds like it would be simple to test using a mixed-effects ANOVA, with a repeat-measures (random) factor of the two outcome measures and a between-subjects factor of group. You'd need to do some sort of post-hoc or planned comparison aswell.

You could also do some sort of mixed-effect regression which should get the a similar (or exactly the same result), although this test doesn't really get taught until more "advanced" stats classes normally.
Reply 5
Original post by iammichealjackson
I'd be happy to help another twin peaks fan! Firstly, ANOVA and multiple regression are essentially the same test some people just argue that ANOVA is just a special case of regression (e.g. http://www.theanalysisfactor.com/why-anova-and-linear-regression-are-the-same-analysis/). I don't think you can have continuous predictor variables in ANOVA, which makes it limited compared to regression.

Secondly, likert scales work equally well for ANOVA and regression. If you have a decent (>20-40 sample size) and a decent number of levels on the likert scale (>3-4) then central limit theorem says that you can assume that the sampling distribution outcome variable will be normal (which is what matters).

Thirdly, this sounds like it would be simple to test using a mixed-effects ANOVA, with a repeat-measures (random) factor of the two outcome measures and a between-subjects factor of group. You'd need to do some sort of post-hoc or planned comparison aswell.

You could also do some sort of mixed-effect regression which should get the a similar (or exactly the same result), although this test doesn't really get taught until more "advanced" stats classes normally.


Haha thanks very much for offering your help!

Okay I see, I've often heard that they are the same, but I've always thought that an ANOVA was used more for experiments, and a regression for naturalistic, survey type designs. Out of curiosity, why do you think that is?

Oh that's a relief than, thanks!

Okay, that makes sense. So for an ANOVA, I'd have my pre and post measures for the repeated measures factor, and then the three conditions for the between-subjects factors. But how would I incorporate the individualism to test for an interaction with ANOVA?

I'd much rather conduct an ANOVA than a regression as I've never even conducted a multiple regression, but my supervisor seems much more keen on regression, I think that's what he tends to use as he conducts a lot of surveys.

Again, thank you so much for your time! I'd love to have your wisdom in this area, some day maybe...Although I can't ever imagine shaking off the fear of stats!
(edited 8 years ago)
Original post by Twinpeaks
Haha thanks very much for offering your help!

Okay I see, I've often heard that they are the same, but I've always thought that an ANOVA was used more for experiments, and a regression for naturalistic, survey type designs. Out of curiosity, why do you think that is?

Oh that's a relief than, thanks!

Okay, that makes sense. So for an ANOVA, I'd have my pre and post measures for the repeated measures factor, and then the three conditions for the between-subjects factors. But how would I incorporate the individualism to test for an interaction with ANOVA?

I'd much rather conduct an ANOVA than a regression as I've never even conducted a multiple regression, but my supervisor seems much more keen on regression, I think that's what he tends to use as he conducts a lot of surveys.

Again, thank you so much for your time! I'd love to have your wisdom in this area, some day maybe...Although I can't ever imagine shaking off the fear of stats!


Hmmm. Well you've made life abit difficult by having three levels on the group factor and a continuous moderating variable. Its definitely doable but i'm not sure how to run the analysis of the top of my head. Maybe it could be done using ANCOVA but i'm not too familiar with that either.

I think the difference is partly historical and also practical (i.e. when people did statistics by hand the ANOVA calculations may be easier when you have discrete groups and you only need to work out means and SDs). There's a very good intro to the history of these tests in Cohen's Applied multple regression book: (p.4; MRC means multiple regression):

1.2 A COMPARISON OF MULTIPLE REGRESSION/CORRELATION AND ANALYSIS OF VARIANCEAPPROACHES

MRC, ANOVA, and ANCOVA are each special cases of the general linear model in mathe matical statistics.2 The description of MRC in this book includes extensions of conventional MRC analysis to the point where it is essentially equivalent to the general linear model. It thus follows that any data analyzable by ANOVA/ANCOVA may be analyzed by MRC, whereas the reverse is not the case. For example, research designs that study how a scaled characteristic of participants (e.g., IQ) and an experimental manipulation (e.g., structured vs. unstructured tasks) jointly influence the subjects' responses (e.g., task performance) cannot readily be fit into the ANOVA framework. Even experiments with factorial designs with unequal cell sam ple sizes present complexities for ANOVAapproaches because of the nonindependence of the factors, and standard computer programs now use a regression approach to estimate effects in such cases. The latter chapters of the book will extend the basic MRC model still further to include alternative statistical methods of estimating relationships.

1.2.1 Historical Background

Historically, MRC arose in the biological and behavioral sciences around 1900 in the study of the natural covariation of observed characteristics of samples of subjects, including Gallon's studies of the relationship between the heights of fathers and sons and Pearson's and Yule's work on educational issues (Yule, 1911). Somewhat later, ANOVA/ANCOVA grew out of the analysis of agricultural data produced by the controlled variation of treatment conditions in manipulative experiments. It is noteworthy that Fisher's initial statistical work in this area emphasized the multiple regression framework because of its generality (see Tatsuoka, 1993). However, multiple regression was often computationally intractable in the precomputer era: computations that take milliseconds by computer required weeks or even months to do by hand. This led Fisher to develop the computationally simpler, equal (or proportional) sample size ANOVA/ANCOVA model, which is particularly applicable to planned experiments. Thus multiple regression and ANOVA/ANCOVAapproaches developed in parallel and, from the perspective of the substantive researchers who used them, largely independently. Indeed, in certain disciplines such as psychology and education, the association of MRC with nonexper imental, observational, and survey research led some scientists to perceive MRC to be less scientifically respectable than ANOVA/ANCOVA, which was associated with experiments. Close examination suggests that this guilt (or virtue) by association is unwarranted—the result of the confusion of data-analytic method with the logical considerations that govern the inference of causality. Experiments in which different treatments are applied to randomly assigned groups of subjects and there is no loss (attrition) of subjects permit unambiguous inference of causality; the observation of associations among variables in a group of ran domly selected subjects does not. Thus, interpretation of a finding of superior early school achievement of children who participate in Head Start programs compared to nonparticipating children depends on the design of the investigation (Shadish, Cook, & Campbell, 2002; West, Biesanz, & Pitts, 2000). For the investigator who randomly assigns children to Head Start versus Control programs, attribution of the effect to program content is straightforward. For the investigator who simply observes whether children whose parents select Head Start pro grams have higher school achievement than those who do not, causal inference becomes less certain. Many other possible differences (e.g., child IQ; parent education) may exist betweenthe two groups of children that could potentially account for any findings. But each of the investigative teams may analyze their data using either ANOVA (or equivalently a t test of the mean difference in school achievement) or MRC (a simple one-predictor regression analysis of school achievement as a function of Head Start attendance with its identical t test). The logical status of causal inference is a function of how the data were produced, not how they were analyzed (see further discussion in several chapters, especially in Chapter 12).



The weird thing about regression is that software such as SPSS does not automatically look at interactions. Off the top of my head i could tell you how to do this with a binary(two-group) variable or a continuous, but im not sure when you have three levels on the group. NOrmally this involves centering continious variables or recoding group variables to values (0,1) and mutplying two variables together to make a dummy variable. Try google searching (running interaction analyses in regression SPSS ) something along those lines. Otherwise there are loads of books on regression which should have instructions about how to do this... (the Applied Multiple Regression Book by Cohen is decent, there should be a pdf version on google somewhere)
(edited 8 years ago)

Quick Reply

Latest