[R-meta] When should I use metaregression without interception?

[Michael Dewey]

Dear Martin

If you remove the intercept and you have a single covariate (xxx in your 
example) you are stating that you know for a fact that the estimated 
effect size is zero if the covariate is zero. This is a strong 
assumption. The two models are fundamentally different so it is not 
surprising that the precision of the estimate for xx changes.

Michael

On 08/11/2019 15:11, Martin Lobo wrote:
> I need help you !!!
> 
> 
> Hi. I don't know when I should use the model without the intercept. In my data the significance changes
> 
> Mixed-Effects Model (k = 8; tau^2 estimator: DL)
> 
> tau^2 (estimated amount of residual heterogeneity):     1.0260 (SE = 3.1913)
> tau (square root of estimated tau^2 value):             1.0129
> I^2 (residual heterogeneity / unaccounted variability): 19.10%
> H^2 (unaccounted variability / sampling variability):   1.24
> R^2 (amount of heterogeneity accounted for):            0.00%
> 
> Test for Residual Heterogeneity:
> QE(df = 6) = 7.4162, p-val = 0.2841
> 
> Test of Moderators (coefficient 2):
> QM(df = 1) = 0.0004, p-val = 0.9841
> 
> Model Results:
> 
>           estimate      se     zval    pval    ci.lb    ci.ub
> intrcpt   -3.7436  1.4907  -2.5113  0.0120  -6.6653  -0.8218  *
> xxx        0.0007  0.0354   0.0200  0.9841  -0.0687   0.0702
> 
> ---
> Signif. codes:  0 �***� 0.001 �**� 0.01 �*� 0.05 �.� 0.1 � � 1
> 
>> metareg(MetaTAV, xxx, intercept = F)
> 
> Mixed-Effects Model (k = 8; tau^2 estimator: DL)
> 
> tau^2 (estimated amount of residual heterogeneity):     5.9922 (SE = 6.3402)
> tau (square root of estimated tau^2 value):             2.4479
> I^2 (residual heterogeneity / unaccounted variability): 62.90%
> H^2 (unaccounted variability / sampling variability):   2.70
> 
> Test for Residual Heterogeneity:
> QE(df = 7) = 18.8661, p-val = 0.0086
> 
> Model Results:
> 
>       estimate      se     zval    pval    ci.lb    ci.ub
> xxx   -0.0917  0.0332  -2.7595  0.0058  -0.1568  -0.0266  **
> 
> 
> another question.
> If I want to make a meta-regression with the cholesterol variable and I have the difference in cholesterol in the two branches, should I put the two differences in the model? I don't know how to assemble the model in that case.
> Best
> 
> 
> 
> Lorenzo Mart�n Lobo MTSAC, FACC, FESC
> Especialista Jerarquizado en Cardiolog�a
> Jefe de Cardiolog�a Hospital Militar Campo de Mayo
> Ex Jefe de Unidad Coronaria Hospital Militar Campo de Mayo
> Miembro Titular de la Sociedad Argentina de Cardiolog�a
> Fellow American College of Cardiology
> Fellow European Society of Cardiology
> Miembro del Area de Investigaci�n de la SAC
> Ex Director del Consejo de Aterosclerosis y Trombosis de la SAC
> Miembro Asesor del Consejo de Aterosclerosis y Trombosis de la SAC
> Ex Director del Consejo de Epidemiolog�a y Prevenci�n Cardiovascular de la SAC
> 
> Miembro Asesor del Consejo de Epidemiolog�a y Prevenci�n Cardiovascular de la SAC
> 
> 
> Instructor de ACLS de la American Heart Association
> 
> 
> ________________________________
> De: Viechtbauer, Wolfgang (SP) <wolfgang.viechtbauer using maastrichtuniversity.nl>
> Enviado: jueves, 7 de noviembre de 2019 11:30
> Para: Martin Lobo <mlobo4370 using hotmail.com>; r-sig-meta-analysis using r-project.org <r-sig-meta-analysis using r-project.org>
> Asunto: RE: Bubble plot in regresion whith two variables
> 
> Hi Lorenzo,
> 
> So, if I understand you correctly, you want to show the line for one variable while holding the other variable constant. Here is the same example from the metafor website extended to this case:
> 
> ########################################
> 
> library(metafor)
> 
> dat <- escalc(measure="RR", ai=tpos, bi=tneg, ci=cpos, di=cneg, data=dat.bcg)
> res <- rma(yi, vi, mods = ~ ablat + year, data=dat)
> 
> size <- 1 / sqrt(dat$vi)
> size <- size / max(size)
> 
> plot(NA, NA, xlim=c(10,60), ylim=c(0.2,1.6),
>       xlab="Absolute Latitude", ylab="Risk Ratio",
>       las=1, bty="l", log="y")
> 
> symbols(dat$ablat, exp(dat$yi), circles=size, inches=FALSE, add=TRUE, bg="black")
> 
> preds <- predict(res, newmods=cbind(0:60, 1969), transf=exp)
> lines(0:60, preds$pred)
> 
> abline(h=1, lty="dotted")
> 
> ########################################
> 
> So, here, I plot the line for 'ablat' while holding year constant at 1969 (which is the median value of the year variable).
> 
> Best,
> Wolfgang
> 
> -----Original Message-----
> From: Martin Lobo [mailto:mlobo4370 using hotmail.com]
> Sent: Thursday, 07 November, 2019 15:12
> To: Viechtbauer, Wolfgang (SP); r-sig-meta-analysis using r-project.org
> Subject: RE: Bubble plot in regresion whith two variables
> 
> Thank's Wolfgang.
> 
> The bubble function of the target package only graphs the first variable and does not allow adding the adjustment line of the multivariate model. I don't know if it explains well to me, I need to add the model adjustment line with two variables. The example you have given me has not been able to adapt it to work with my data.
> Thank you
> 
> Lorenzo Mart�n Lobo MTSAC, FACC, FESC
> 
> 	[[alternative HTML version deleted]]
> 
> 
> 
> 
> _______________________________________________
> R-sig-meta-analysis mailing list
> R-sig-meta-analysis using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
> 

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