Should I use a 2-way ANCOVA?

[lir6777]

CONTENTS DELETED

The author has deleted this message.

[Art Kendall]

It is great that you are seeking advice before doing the study.

it seems that you want to look at some form of differences in change model. There are varying opinions these days about how to measure change. In strict terminology an ANCOVA would not include the interaction of the pretest with the other predictors, so I would suggest a more inclusive form of the General Linear Model.
You would want to do 1) repeated measures and/or 2) include the pretest as a predictor and include the pretest by other predictors interaction in the model.

In order to help the people on the list would need more information.

To start (others may ask additional questions before) this is wrapped up.

Why are you limiting the study to 2 levels of disability?
Why are you limiting the study to two levels of blame?

I suggest that you leave alpha at the conventional .05 and power at .80 and determine your study size according to the effect size that means something.
What size  difference on the DV would cause you to reject the default position in policy/practice/theory?

Will you be able to test all 4 conditions in each school?
Why are you excluding school as a predictor (IV)?

How will you know that observed differences were not due to different speakers and/or environments in which the talk is given?

How will you recruit schools?
How will you recruit students?


I suggest that you reply to the list so that others can contribute and/or learn from the conversation.

Art Kendall
Social Research Consultants

On 10/24/2011 7:20 AM, lir6777 wrote:

Hi,
I'm not sure if this is the right place but I'm looking for some advice on
stats and you guys seem to know your stuff!
I'm fairly new in my job and putting together a research study so I want to
make sure I'm on the right path so as not to look a dope in front of my new
boss!
Okay, so my study focuses on stigma of disability, and will look like this:

DV- stigmatising attitudes (measured by established questionnaire,
continuous variable)
IVs- 1) level of disability- dichotomous variable i.e. uses wheelchair (call
this 1A) Vs does not need wheelchair (1B)
2) blame- dichotomous, person seen as having caused their disability (2A) Vs
did not cause disability (2B).

So I will have four conditions- students will receive a talk from one of
four people -

1a/2a, 1a/2b, 1b/2a, or 1b/2b.

Students' stigmatising attitudes will be measured pre- and post-talk.

So what I want to look at is what the main effects of the two IVs are, as
well as interaction. Because different schools may show different attitudes
to start with (depending on socio-economic factors etc) I thought pre-talk
scores should be a covariate.

Can I then use a 2-way ANCOVA?

And how would I find out how many participants I will need given the number
of conditions?

I'd appreciate some expert advice- my stats classes in college seem a long
time ago now!
Thanks!

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Should-I-use-a-2-way-ANCOVA-tp4932158p4932158.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[lir6777]

Hi Art,
Thanks so much for your response.
Okay so to answer your questions:
Only 2 levels of disability because past research has shown that a visible physical impairment in general reduces stigma when compared to more 'hidden' cognitive/emotional/behavioural signs of disability. So to keep it simple, I chose visible disability vs no visible disability.
2 levels of blame- this is based on a previous study which broke 5 aetiologies into 'blame' and 'no-blame' categories, and found that 'blame' categories were more stigmatised.So I am keeping with this simplicity and wish to avoid less straightforward aetiologies (for example, a person who got drunk, stood in the middle of the road and got hit by a car- this could be seen as his fault or the driver's fault- too messy for this study!)

(Both of these could be investigated further in future studies; as this is an initial investigation, I was trying to keep it as simple as possible- not sure if I've managed that!)

Testing all 4 conditions at each school would not be possible-  it would involve huge input of time from our client participants who will be giving the talks to students about their experience of their disability, and would either involve repeated measures or a lot of free classrooms in each school!
School as a predictor- hadn't thought about it if I'm honest. I haven't approached schools yet so the possible variance isn't too clear yet..There also isn't research to base this on, apart from that on socio-economic status- but making an assumption of SES by school isn't sensible.

Of course I won't be able to know for sure that speakers' individual personalities and the environment of talk haven't impacted the results- I will try to match the speakers as closely as possible in terms of personality etc., and ask schools for an empty room for the talk, rather than a lunchroom etc. These will be unavoidable limitations as far as I can see?

I will make a list of all secondary schools in the region and contact all of them in writing with information sheet and invitation for 4th years to participate. I will then follow-up with phonecall 2 weeks later.
Participating schools then- Students will be given the information by their teacher/headmaster, and be given an information sheet and consent form for their parents. They and their parents can then opt in or out.

Any more questions or if I'm not clear please ask- I'm really glad to have some expert input!
Thanks!


Lynn Irwin

Assistant Psychologist
Acquired Brain Injury Ireland

Ballymaley Business Park, Gort Road, Ennis, Co. Clare.
Tel.: [phone]                           Mobile: [phone]
________________________________________
From: Art Kendall [via SPSSX Discussion] [[hidden email]]
Sent: Monday, October 24, 2011 3:46 PM
To: Lynn Irwin
Subject: Re: Should I use a 2-way ANCOVA?

It is great that you are seeking advice before doing the study.

it seems that you want to look at some form of differences in change model. There are varying opinions these days about how to measure change. In strict terminology an ANCOVA would not include the interaction of the pretest with the other predictors, so I would suggest a more inclusive form of the General Linear Model.
You would want to do 1) repeated measures and/or 2) include the pretest as a predictor and include the pretest by other predictors interaction in the model.

In order to help the people on the list would need more information.

To start (others may ask additional questions before) this is wrapped up.

Why are you limiting the study to 2 levels of disability?
Why are you limiting the study to two levels of blame?

I suggest that you leave alpha at the conventional .05 and power at .80 and determine your study size according to the effect size that means something.
What size  difference on the DV would cause you to reject the default position in policy/practice/theory?

Will you be able to test all 4 conditions in each school?
Why are you excluding school as a predictor (IV)?

How will you know that observed differences were not due to different speakers and/or environments in which the talk is given?

How will you recruit schools?
How will you recruit students?


I suggest that you reply to the list so that others can contribute and/or learn from the conversation.

Art Kendall
Social Research Consultants

On 10/24/2011 7:20 AM, lir6777 wrote:

Hi,
I'm not sure if this is the right place but I'm looking for some advice on
stats and you guys seem to know your stuff!
I'm fairly new in my job and putting together a research study so I want to
make sure I'm on the right path so as not to look a dope in front of my new
boss!
Okay, so my study focuses on stigma of disability, and will look like this:

DV- stigmatising attitudes (measured by established questionnaire,
continuous variable)
IVs- 1) level of disability- dichotomous variable i.e. uses wheelchair (call
this 1A) Vs does not need wheelchair (1B)
2) blame- dichotomous, person seen as having caused their disability (2A) Vs
did not cause disability (2B).

So I will have four conditions- students will receive a talk from one of
four people -

1a/2a, 1a/2b, 1b/2a, or 1b/2b.

Students' stigmatising attitudes will be measured pre- and post-talk.

So what I want to look at is what the main effects of the two IVs are, as
well as interaction. Because different schools may show different attitudes
to start with (depending on socio-economic factors etc) I thought pre-talk
scores should be a covariate.

Can I then use a 2-way ANCOVA?

And how would I find out how many participants I will need given the number
of conditions?

I'd appreciate some expert advice- my stats classes in college seem a long
time ago now!
Thanks!

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Should-I-use-a-2-way-ANCOVA-tp4932158p4932158.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email]</user/SendEmail.jtp?type=node&node=4932863&i=0> (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD



===================== To manage your subscription to SPSSX-L, send a message to [hidden email]</user/SendEmail.jtp?type=node&node=4932863&i=1> (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

________________________________
If you reply to this email, your message will be added to the discussion below:
http://spssx-discussion.1045642.n5.nabble.com/Should-I-use-a-2-way-ANCOVA-tp4932158p4932863.html
To unsubscribe from Should I use a 2-way ANCOVA?, click here<

ABI & The Carer - An Information & Education Event for Family Carers & Healthcare Professionals - 17th November Citywest Hotel, Saggart.
http://www.abiireland.ie/events.html

Winner of the Astellas Changing Tomorrow Awards 2010 - Hope Category