Message-ID: <964474a4-c451-8757-8609-2a0c1ecf7899@dewey.myzen.co.uk>
Date: 2023-12-12T13:26:22Z
From: Michael Dewey
Subject: [R-meta] Follow-up the coding question for meta-regression
In-Reply-To: <CAPm5Ok33gF4zM=_OUw3ASvvEZzX0eNZcB0F8BNmUTjDHRh_HRQ@mail.gmail.com>
Comments in-line
On 11/12/2023 23:21, ?????? via R-sig-meta-analysis wrote:
> Dear Wolfgang,
> Sorry for another interruption. I ran into two more questions. First,
> when I was trying to write the code like this:
>
> dat$duration <- cut(dat$duration, breaks=c(0,4,8,12),
> labels=c(?NA?,?S?,?M?,?L?), right =FALSE)
>
> and the system told me that ?x has to be numbers?, what should I do then?
> (p.s. I am pretty sure that the part ?breaks? is also wrong?)
>
> **Name of the moderator: duration
> The distinction between the moderators:
> NA = not provided
> S (short) = 0-4 weeks
> M (medium) = 4-8 weeks
> L (long) = 8 weeks or above
So R is correct, that variable is not numeric but either character or
factor. You would not need breaks either.
>
> and my second question is that, here is one of the interaction I
> performed:
> [here I was trying to find out the interaction between the moderator
> "instruction type" and "research design", and instruction type was coded
> "ins" and with two categorical value "D (definition)" and "C(context)";
> research design was coded "rede" and with two categorical value "D
> (traditional)" and "O (others)"]
>
> Model Results:
> estimate se zval pval ci.lb ci.ub
> intrcpt 0.9482 0.3581 2.6476 0.0081 0.2463 1.6501 **
> insD 0.5142 0.4281 1.2012 0.2297 -0.3248 1.3532
> redeO 0.4920 0.6163 0.7984 0.4246 -0.7158 1.6998
> insD:redeO -1.2206 0.8256 -1.4784 0.1393 -2.8388 0.3976
>
>
> Does the last line "insD:redeO" already means the interaction was not
> significant with the p-value of 0.1393? Do I need to do the "anova(res.il,
> btt3:4)" as suggested in the website?
Yes, that interaction is not adding anything important in your model.
Michael
>
> Nick
> ??? (Nick Chen)
> Email: t571 at wlgsh.tp.edu.tw <t5741 at wlgsh.tp.edu.tw>
>
> [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-meta-analysis mailing list @ R-sig-meta-analysis at r-project.org
> To manage your subscription to this mailing list, go to:
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>
--
Michael