[R-meta] Fwd: About the coefficient in meta-regression analysis

[Michael Dewey]

Dear Nick

Comments in-line

On 19/12/2023 14:31, 英文科陳品誠 via R-sig-meta-analysis wrote:
> Dear Michael and all,
> 
>         Thank you for the swift response. I do have some further questions
> to ask.
>         (a) So, in order to "remove" the intercept, what code should I use?

I do not use the same package as you but where you currently have educ 
as the moderator try putting educ - 1

>         (b) Let's say my research is about whether using digital games will
> affect the result of English vocabulary acquisition or not. So as you said,
> "... remove the intercept so each coefficient would then be tested for its
> difference from zero." , does that mean if I rerun the code again with the
> removal of intercept and presumably get an estimate of 0.3 from the
> subgroup "eduH" and the p-value is less than 0.05. Can I say that "eduH has
> a positive coefficient to English vocabulary acquisition"? If not, what
> would the coefficient mean here (since there is no intercept to compare to)?

It is comparing the estimated effect size with zero. Whether that is a 
meaningful scientific question is up to you. I would be surprised if the 
coefficient did turn out to be 0.3 but try it and see. Note that the 
overall test for the effect of educ will be testing a different 
hypothesis when you remove the intercept.

Michael

> 
>       Thanks again for your kindly help!
>      Nick
> 陳品誠 (Nick Chen)
> Email: t571 using wlgsh.tp.edu.tw <t5741 using wlgsh.tp.edu.tw>
> 
> 
> Michael Dewey <lists using dewey.myzen.co.uk> 於 2023年12月19日 週二 下午9:14寫道：
> 
>> Dear Nick
>>
>> Comments in line
>>
>> On 19/12/2023 11:51, 英文科陳品誠 via R-sig-meta-analysis wrote:
>>> Dear all,
>>>          This is part of my meta-regression analysis result after I use
>> the
>>> following code:
>>>
>>>> m.gen.edu <- metareg(m.gen, ~edu)
>>>
>>>       Model results:
>>>                                  estimate  .....
>>> intercept                0.6820   .....
>>> eduH                      0.4590   ......
>>> eduI                        -0.0834 ......
>>> eduK                       0.5668 ........
>>> eduS                       0.2473   .......
>>>
>>>        The *edu *here stands for the moderator "educational level" from my
>>> dataset and the *eduH, eduI, eduK*, and *eduS* stands for different
>>> subgroups. Does the 0.4590 here represent the coefficient of the subgroup
>>> *eduH* ?
>>
>> In a model with an intercept each level is compared to a reference level
>> so the value, 0.4590, is the difference estimated between group H and
>> the reference group.
>>
>>    And is there a code to get the overall coefficient of the
>>> moderator "educational level" rather than the individual coefficient of
>>> each subgroup?
>>
>> The short answer is "No". You can do a test for the overall effect of
>> educational level. You could also in a simple model like this remove the
>> intercept so each coefficient would then be tested for its difference
>> from zero.
>>
>> Michael
>>
>>>
>>>    Nick Chen
>>>
>>> 陳品誠 (Nick Chen)
>>> Email: t571 using wlgsh.tp.edu.tw <t5741 using wlgsh.tp.edu.tw>
>>>
>>>        [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-sig-meta-analysis mailing list @ R-sig-meta-analysis using r-project.org
>>> To manage your subscription to this mailing list, go to:
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>>
>>
>> --
>> Michael
>>
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-meta-analysis mailing list @ R-sig-meta-analysis using r-project.org
> To manage your subscription to this mailing list, go to:
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
> 

-- 
Michael