Mediation Bootstrap results

[OLMABA JALA]

Dear all,

I am running a mediation in Stata and I am getting significant effects of the mediator and the IV
sureg (mv1 iv)(dv mv1 iv)
After that I want to bootstrap the results

bootstrap r(ind_eff) r(dir_eff), reps(5000)

But now the indirect effect _bs_1 has a p value of 0.3 and thereby is not significant anymore

Could you please help me how I can interpret the results?
Is the model with a mediation not working for me?

when running estat bootstrap, percentile the interval value is not 0

[Felix Bittmann]

From your example it is not fully clear to me what you are doing and which programs you use. I assume, some of them are written by you and not Stata standard. So it would be interesting to know these in detail. For example, the command you show is not complete since the part after the colon is missing. Which program are you bootstrapping?

[OLMABA JALA]

Thank you Felix

I am just using STATA programs

What I would like to do: I would like to check if there is a mediation of mv on the relationship between the IV and the DV
So first I am running this code -sureg (mv1 iv)(dv mv1 iv)-

I am bootstrapping the mediation-sureg (mv1 iv)(dv mv1 iv)-

[OLMABA JALA]

Does anyone have an idea?

My STATA commands are
sureg (mv1 iv) (dv mv1 iv)
bootstrap r(ind_eff) r(dir_eff), reps(2000): sureg (mv1 iv) (dv mv1 iv)

I am getting significant effects of the mediator and the IV for the first commands
But after the second command the indirect effect _bs_1 has a p value of 0.3 and thereby is not significant anymore

Could you please help me how I can interpret the results?
Is the model with a mediation not working for me?

when running estat bootstrap, percentile the interval value is not 0

[Felix Bittmann]

I still cannot replicate your commands. Normally, sureg does not compute the statistics you specify. Here is a simple example (Stata 16.1):

. sysuse nlsw88, clear
(NLSW, 1988 extract)

. sureg (hours age collgrad) (wage hours age collgrad)

Seemingly unrelated regression
--------------------------------------------------------------------------
Equation             Obs   Parms        RMSE    "R-sq"       chi2        P
--------------------------------------------------------------------------
hours              2,242       2    10.46462    0.0080      18.11   0.0001
wage               2,242       3    5.487714    0.0911     224.80   0.0000
--------------------------------------------------------------------------

------------------------------------------------------------------------------
             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
hours        |
         age |  -.0932333   .0722292    -1.29   0.197    -.2347999    .0483334
    collgrad |   2.102342   .5198566     4.04   0.000     1.083442    3.121243
       _cons |   40.37051    2.84026    14.21   0.000      34.8037    45.93731
-------------+----------------------------------------------------------------
wage         |
       hours |    .074792   .0110752     6.75   0.000      .053085    .0964989
         age |  -.0562952   .0378915    -1.49   0.137    -.1305612    .0179709
    collgrad |   3.459765   .2736087    12.64   0.000     2.923502    3.996028
       _cons |   6.375591   1.555111     4.10   0.000     3.327629    9.423554
------------------------------------------------------------------------------

. bootstrap r(ind_eff) r(dir_eff), reps(999) seed(123): ///
>         sureg (hours age collgrad) (wage hours age collgrad)
(running sureg on estimation sample)
'r(ind_eff)' evaluated to missing in full sample
r(322);

Therefore, I assume you must have some ado or something you are using. The regular sureg command does not return these statistics.

[OLMABA JALA]

Thanks
I am not aware of using any other packages

But I tried the same using the approach from here: https://stats.idre.ucla.edu/stata/fa...data-method-1/
bootstrap indeff=r(ind_eff) direff=r(dir_eff) toteff=r(tot_eff), /// reps(500) seed(1) cluster(cid) idcluster(ncid): /// ml_mediation, dv(write) iv(hon) mv(abil) l2id(ncid)

[OLMABA JALA]

Does anyone have an idea?

[Phillip Spaeth]

Hi all,

New to Stata and to Statalist. I have installed the ml_mediation package and am following the syntax here.

The following syntax is working fine:
ml_mediation, dv(PSWQ) iv(time) mv(Attention) l2id(ID)
return list

However, when I enter the bootstrapping syntax...
bootstrap indeff=r(ind_eff) direff=r(dir_eff) toteff=r(tot_eff), ///
reps(5000) seed(1) cluster(ID) idcluster(nID): ///
ml_mediation, dv(PSWQ) iv(time) mv(Attention) l2id(nID)
estat boot, percentile bc
...I get the following errors:

. bootstrap indeff=r(ind_eff) direff=r(dir_eff) toteff=r(tot_eff), ///
indeff=r(ind_eff) command not found
r(111);

.
. reps(5000) seed(1) cluster(ID) idcluster(nID): ///
command reps is unrecognized
r(199);

Does anyone have any ideas why I'm getting errors for these commands?

[Carlo Lazzaro]

Obama/Phillip:
what if:

bootstrap indeff=r(ind_eff) direff=r(dir_eff) toteff=r(tot_eff), reps(500) seed(1) cluster(cid) idcluster(ncid): ml_mediation, dv(write) iv(hon) mv(abil) l2id(ncid)

[Phillip Spaeth]

That did the trick, Carlo! Thank you so much for your help.