[R-meta] Moderator level getting removed from model

[Michael Dewey]

Dear Shannon

Perhaps I misunderstand you but why not just put a footnote in your 
table explaining why there is no estimate? You might want to clarify 
there whether this zero is a structural zero (in principle is zero) or a 
sampling zero (could have been positive but in your data is zero).

Michael

On 12/02/2018 21:08, Shannon Duncan wrote:
> Hi Wolfgang,
> 
> The table does reveal that there are no observations for the "1" coded in
> Decision Aids for the 2 tridiff level.
> 
> We're essentially yes looking for a way to show this is in the reporting of
> the meta analysis.
> 
> Thank!
> Shannon
> 
>       0  1
>    1 32 12
>    2 10  0
>    3 11 12
> 
> 
> 
> On Mon, Feb 12, 2018 at 12:56 PM, Viechtbauer Wolfgang (SP) <
> wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:
> 
>> Dear Shannon,
>>
>> Indeed, rma() will omit a coefficient from the model if it is not
>> estimable. What does
>>
>> table(choiceaccdata$tridiff, choiceaccdata$Provides.Decision.Aids)
>>
>> show? I suspect this will reveal why this coefficient is not estimable.
>>
>> So the only thing one could print for that coefficient would be NA (not
>> available). Are you looking for a way to show this?
>>
>> Best,
>> Wolfgang
>>
>>> -----Original Message-----
>>> From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-
>>> project.org] On Behalf Of Shannon Duncan
>>> Sent: Monday, 12 February, 2018 18:23
>>> To: r-sig-meta-analysis at r-project.org
>>> Subject: [R-meta] Moderator level getting removed from model
>>>
>>> Hello,
>>>
>>> I have a question that I can't seem to find an answer to, hoping someone
>>> here can help me out. We are running the following model:
>>>
>>> resalt <- metafor::rma(cohens_d, variance, mods = ~ factor(tridiff) *
>>> Provides.Decision.Aids, test = "t", data = choiceaccdata)
>>>
>>> resalt
>>>
>>> ## tau^2 (estimated amount of residual heterogeneity):     0.0574 (SE =
>>> 0.0175)
>>> ## tau (square root of estimated tau^2 value):             0.2395
>>> ## I^2 (residual heterogeneity / unaccounted variability): 67.78%
>>> ## H^2 (unaccounted variability / sampling variability):   3.10
>>> ## R^2 (amount of heterogeneity accounted for):            39.40%
>>> ##
>>> ## Test for Residual Heterogeneity:
>>> ## QE(df = 72) = 195.8927, p-val < .0001
>>> ##
>>> ## Test of Moderators (coefficient(s) 2:5):
>>> ## F(df1 = 4, df2 = 72) = 6.8504, p-val < .0001
>>> ##
>>> ## Model Results:
>>> ##
>>> ##                                          estimate      se     tval
>>> pval
>>> ## intrcpt                                   -0.2234  0.0527  -4.2434
>>> <.0001
>>> ## factor(tridiff)2                          -0.1642  0.1415  -1.1609
>>> 0.2495
>>> ## factor(tridiff)3                          -0.4823  0.1523  -3.1656
>>> 0.0023
>>> ## Provides.Decision.Aids                     0.2183  0.0996   2.1910
>>> 0.0317
>>> ## factor(tridiff)3:Provides.Decision.Aids    0.6595  0.2218   2.9736
>>> 0.0040
>>> ##                                            ci.lb    ci.ub
>>> ## intrcpt                                  -0.3284  -0.1185  ***
>>> ## factor(tridiff)2                         -0.4462   0.1178
>>> ## factor(tridiff)3                         -0.7860  -0.1786   **
>>> ## Provides.Decision.Aids                    0.0197   0.4169    *
>>> ## factor(tridiff)3:Provides.Decision.Aids   0.2174   1.1017   **
>>>
>>> As you can see, it's not showing the second level of the factor  tridiff
>>> (tridiff2) interacting with Decision Aids. I realise this is probably due
>>> to redundancy or something of that sort, but we would like to see it
>>> printed out. Is that at all possible?
>>>
>>> Thank you for your time and any insight you can provide,
>>> Shannon
>>
> 
> 
> 

-- 
Michael
http://www.dewey.myzen.co.uk/home.html