f

#### Should I use Two-way Repeated ANOVA for this?

```I want to analyze whether a psychological treatment is effective in reducing patients' anxiety toward cancer assessment at different time points (before, during, and after).

There will be 2 groups, Intervention and Control, each consists of similar number of participants.

So the data will look like this:

Groups\ Time	T1	T2	T3

Intervention	I1	I2	I3
Control	        C1	C2	C3

1) Should I use Two-way Repeated ANOVA?

Besides anxiety, I also hypothesized that patients' educational level might also affect their anxiety. Should I use Pearson's Correlation to examine the strength and significance before using it as a Covariate (or is there any other steps I should do)?

Thanks.
```
 0
fairydust
12/5/2016 9:28:17 AM
comp.soft-sys.stat.spss 5679 articles. 0 followers.

3 Replies
611 Views

Similar Articles

[PageSpeed] 51

```On Monday, December 5, 2016 at 4:28:18 AM UTC-5, fairydust wrote:
> I want to analyze whether a psychological treatment is effective in reduc=
ing patients' anxiety toward cancer assessment at different time points (be=
fore, during, and after).
>=20
> There will be 2 groups, Intervention and Control, each consists of simila=
r number of participants.=20
>=20
> So the data will look like this:
>=20
> Groups\ Time	T1	T2	T3
>=20
> Intervention	I1	I2	I3
> Control	        C1	C2	C3
>=20
>=20
>=20
> 1) Should I use Two-way Repeated ANOVA?
>=20
>=20
> Besides anxiety, I also hypothesized that patients' educational level mig=
ht also affect their anxiety. Should I use Pearson's Correlation to examine=
the strength and significance before using it as a Covariate (or is there =
any other steps I should do)?
>=20
> Thanks.

The T1 measure of anxiety is a baseline measure, by the sounds of it. So yo=
u could treat it as a covariate in a 2x2 repeated measures ANCOVA.  If you =
take that approach, you might find it useful to center the covariate.  See =
the following letter to the editor, for example.

http://www.nmr.mgh.harvard.edu/~kdijk/PDF/Van_Breukelen_and_Van_Dijk_2007_J=
INS.pdf

HTH.
```
 0
Bruce
12/5/2016 6:59:08 PM
```On Tuesday, December 6, 2016 at 2:59:10 AM UTC+8, Bruce Weaver wrote:
> On Monday, December 5, 2016 at 4:28:18 AM UTC-5, fairydust wrote:
> > I want to analyze whether a psychological treatment is effective in red=
ucing patients' anxiety toward cancer assessment at different time points (=
before, during, and after).
> >=20
> > There will be 2 groups, Intervention and Control, each consists of simi=
lar number of participants.=20
> >=20
> > So the data will look like this:
> >=20
> > Groups\ Time	T1	T2	T3
> >=20
> > Intervention	I1	I2	I3
> > Control	        C1	C2	C3
> >=20
> >=20
> >=20
> > 1) Should I use Two-way Repeated ANOVA?
> >=20
> >=20
> > Besides anxiety, I also hypothesized that patients' educational level m=
ight also affect their anxiety. Should I use Pearson's Correlation to exami=
ne the strength and significance before using it as a Covariate (or is ther=
e any other steps I should do)?
> >=20
> > Thanks.
>=20
> The T1 measure of anxiety is a baseline measure, by the sounds of it. So =
you could treat it as a covariate in a 2x2 repeated measures ANCOVA.  If yo=
u take that approach, you might find it useful to center the covariate.  Se=
e the following letter to the editor, for example.
>=20
> http://www.nmr.mgh.harvard.edu/~kdijk/PDF/Van_Breukelen_and_Van_Dijk_2007=
_JINS.pdf
>=20
> HTH.

Thanks for the reply, Bruce. One of my colleagues thinks that it's a bit ri=
sky calling my research an RCT (for its not-so-stringent sequence of group =
assignment; so it should not be called "randomized"). In this case, does th=
e 2x2 repeated measures ANCOVA still apply?
```
 0
fairydust
12/6/2016 2:19:21 AM
```On Mon, 5 Dec 2016 18:19:21 -0800 (PST), fairydust
<kayingchau92@gmail.com> wrote:

>On Tuesday, December 6, 2016 at 2:59:10 AM UTC+8, Bruce Weaver wrote:
>> On Monday, December 5, 2016 at 4:28:18 AM UTC-5, fairydust wrote:
>> > I want to analyze whether a psychological treatment is effective in reducing patients' anxiety toward cancer assessment at different time points (before, during, and after).
>> >
>> > There will be 2 groups, Intervention and Control, each consists of similar number of participants.
>> >
>> > So the data will look like this:
>> >
>> > Groups\ Time	T1	T2	T3
>> >
>> > Intervention	I1	I2	I3
>> > Control	        C1	C2	C3
>> >
>> >
>> >
>> > 1) Should I use Two-way Repeated ANOVA?
>> >
>> >
>> > Besides anxiety, I also hypothesized that patients' educational level might also affect their anxiety. Should I use Pearson's Correlation to examine the strength and significance before using it as a Covariate (or is there any other steps I should do)?
>> >
>> > Thanks.
>>
>> The T1 measure of anxiety is a baseline measure, by the sounds of it. So you could treat it as a covariate in a 2x2 repeated measures ANCOVA.  If you take that approach, you might find it useful to center the covariate.  See the following letter to the editor, for example.
>>
>> http://www.nmr.mgh.harvard.edu/~kdijk/PDF/Van_Breukelen_and_Van_Dijk_2007_JINS.pdf
>>
>> HTH.
>
>Thanks for the reply, Bruce. One of my colleagues thinks that it's a bit risky calling my research an RCT (for its not-so-stringent sequence of group assignment; so it should not be called "randomized"). In this case, does the 2x2 repeated measures ANCOVA still apply?

The fact that the study is not randomized does not give you
any new possibility for a test.  What it does, instead, is make
the conclusions iffier.  You need to do what you can, for instance,
to argue that the groups /are/ comparable, especially in the
aspects that might matter for the ANOVA or the conclusions.

In particular, it becomes more likely that there /will/ be a
difference at Pre that you should describe and consider
carefully. Drawing graphs of means can be helpful.

--
Rich Ulrich

```
 0
Rich
12/6/2016 3:11:11 AM