Pre and post treatment

[Stata Beginner]

Hello, I have done RCT random assigned to treatment A and B and follow at 1 and 2 months. I want to analyze whether there is difference between treatment A and B to affect (month1-baseline) and (month2-baseline).

Group baseline month1 month2
A 2 4 8
A 6 5 10
A 9 15 20
B 3 5 7
B 6 8 10

I can't figure how to analyze these by Stata
Thank you

[Joseph Coveney]

As a for-instance, you could do something along the following lines. The null hypothesis you mention is tested in the group × time interaction term. Look at the grp#tim row in the column labeled "H-F" in table showing adjustments for departure from sphericity, below the main ANOVA output.

.ÿ
.ÿversionÿ16.1

.ÿ
.ÿclearÿ*

.ÿ
.ÿinputÿstr1ÿGroupÿbyte(baselineÿmonth1ÿmonth2)

ÿÿÿÿÿÿÿÿÿGroupÿÿbaselineÿÿÿÿmonth1ÿÿÿÿmonth2
ÿÿ1.ÿAÿ2ÿ4ÿ8
ÿÿ2.ÿAÿ6ÿ5ÿ10
ÿÿ3.ÿAÿ9ÿ15ÿ20
ÿÿ4.ÿBÿ3ÿ5ÿ7
ÿÿ5.ÿBÿ6ÿ8ÿ10
ÿÿ6.ÿend

.ÿ
.ÿ*
.ÿ*ÿBeginÿhere
.ÿ*
.ÿencodeÿGroup,ÿgenerate(grp)label(Groups)

.ÿrenameÿbaselineÿout0

.ÿrenpfixÿmonthÿout

.ÿ
.ÿgenerateÿintÿpidÿ=ÿ_n

.ÿquietlyÿreshapeÿlongÿout,ÿi(pid)ÿj(tim)

.ÿ
.ÿanovaÿoutÿgrpÿ/ÿpid|grpÿtimÿgrp#tim,ÿrepeated(tim)

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿ=ÿÿÿÿÿÿÿÿÿ15ÿÿÿÿR-squaredÿÿÿÿÿ=ÿÿ0.9440
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿRootÿMSEÿÿÿÿÿÿ=ÿÿÿÿ1.68874ÿÿÿÿAdjÿR-squaredÿ=ÿÿ0.8694

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿSourceÿ|ÿPartialÿSSÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿÿÿMSÿÿÿÿÿÿÿÿFÿÿÿÿProb>F
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-----------+----------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿModelÿ|ÿÿ288.62222ÿÿÿÿÿÿÿÿÿÿ8ÿÿÿ36.077778ÿÿÿÿÿ12.65ÿÿ0.0031
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿgrpÿ|ÿÿ18.677778ÿÿÿÿÿÿÿÿÿÿ1ÿÿÿ18.677778ÿÿÿÿÿÿ0.32ÿÿ0.6137
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿpid|grpÿ|ÿÿ177.72222ÿÿÿÿÿÿÿÿÿÿ3ÿÿÿ59.240741ÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-----------+----------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿtimÿ|ÿÿ73.688889ÿÿÿÿÿÿÿÿÿÿ2ÿÿÿ36.844444ÿÿÿÿÿ12.92ÿÿ0.0067
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿgrp#timÿ|ÿÿ6.4888889ÿÿÿÿÿÿÿÿÿÿ2ÿÿÿ3.2444444ÿÿÿÿÿÿ1.14ÿÿ0.3812
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿResidualÿ|ÿÿ17.111111ÿÿÿÿÿÿÿÿÿÿ6ÿÿÿ2.8518519ÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-----------+----------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿTotalÿ|ÿÿ305.73333ÿÿÿÿÿÿÿÿÿ14ÿÿÿ21.838095ÿÿ


Between-subjectsÿerrorÿterm:ÿÿpid|grp
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿLevels:ÿÿ5ÿÿÿÿÿÿÿÿÿ(3ÿdf)
ÿÿÿÿÿLowestÿb.s.e.ÿvariable:ÿÿpid
ÿÿÿÿÿCovarianceÿpooledÿover:ÿÿgrpÿÿÿÿÿÿÿ(forÿrepeatedÿvariable)

Repeatedÿvariable:ÿtim
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿHuynh-Feldtÿepsilonÿÿÿÿÿÿÿÿ=ÿÿ0.8140
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿGreenhouse-Geisserÿepsilonÿ=ÿÿ0.5193
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿBox'sÿconservativeÿepsilonÿ=ÿÿ0.5000

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ------------ÿProbÿ>ÿFÿ------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿSourceÿ|ÿÿÿÿÿdfÿÿÿÿÿÿFÿÿÿÿRegularÿÿÿÿH-FÿÿÿÿÿÿG-GÿÿÿÿÿÿBox
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ-----------+----------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿtimÿ|ÿÿÿÿÿÿ2ÿÿÿÿ12.92ÿÿÿ0.0067ÿÿÿ0.0125ÿÿÿ0.0345ÿÿÿ0.0369
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿgrp#timÿ|ÿÿÿÿÿÿ2ÿÿÿÿÿ1.14ÿÿÿ0.3812ÿÿÿ0.3774ÿÿÿ0.3656ÿÿÿ0.3644
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿResidualÿ|ÿÿÿÿÿÿ6
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ----------------------------------------------------------------

.ÿ
.ÿexit

endÿofÿdo-file


.

Is there a reason why you're not using your name? This isn't homework, is it?

[Stata Beginner]

Thanks, Joseph.
I am new here and I mistakenly create my name but I can't edit the name now.
It's an example I try to described my RCT, not a homework.

I just start to use stata so I try hard to figure this line with syntex / | but I can't understand it's meaning.
anova out grp/pid|grp tim grp#tim, repeated(tim)
Could you explain more?

And If I want to make the output like this
"mean difference between group A and B in month 1 compared to baseline is xxxx +- xxxx mg/dL; p=xxxx"
"mean difference between group A and B in month 2 compared to baseline is xxxx +- xxxx mg/dL; p=xxxx"
What statistical test should I use?

Thank you.

[Joseph Coveney]

anova out grp/pid|grp tim grp#tim, repeated(tim)
Could you explain more?

It's in the user's manual entry for anova under the section heading Repeated-measures ANOVA beginning on Page 48

If I want to make the output like this
"mean difference between group A and B in month 1 compared to baseline is xxxx +- xxxx mg/dL; p=xxxx"
"mean difference between group A and B in month 2 compared to baseline is xxxx +- xxxx mg/dL; p=xxxx"
What statistical test should I use?

You could use test or lincom after fitting the model with anova.

help test
help lincom

for more information.

If you used encode to numerically encode your grouping variable as I did above, then the numeric values for Groups A and B would be 1 and 2. So the contrasts for the two simple effects after the output shown above in #2 would be

lincom (2.grp#1.tim - 2.grp#0.tim) - (1.grp#1.tim - 1.grp#0.tim)

and

lincom (2.grp#2.tim - 2.grp#0.tim) - (1.grp#2.tim - 1.grp#0.tim)

.ÿlincomÿ(2.grp#1.timÿ-ÿ2.grp#0.tim)ÿ-ÿ(1.grp#1.timÿ-ÿ1.grp#0.tim)

ÿ(ÿ1)ÿÿ1b.grp#0b.timÿ-ÿ1b.grp#1o.timÿ-ÿ2o.grp#0b.timÿ+ÿ2.grp#1.timÿ=ÿ0

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿoutÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿ(1)ÿ|ÿÿ-.3333333ÿÿÿ2.180157ÿÿÿÿ-0.15ÿÿÿ0.883ÿÿÿÿ-5.667986ÿÿÿÿÿ5.00132
------------------------------------------------------------------------------

.ÿlincomÿ(2.grp#2.timÿ-ÿ2.grp#0.tim)ÿ-ÿ(1.grp#2.timÿ-ÿ1.grp#0.tim)

ÿ(ÿ1)ÿÿ1b.grp#0b.timÿ-ÿ1b.grp#2o.timÿ-ÿ2o.grp#0b.timÿ+ÿ2.grp#2.timÿ=ÿ0

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿoutÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿ(1)ÿ|ÿÿÿÿÿÿÿÿÿ-3ÿÿÿ2.180157ÿÿÿÿ-1.38ÿÿÿ0.218ÿÿÿÿ-8.334653ÿÿÿÿ2.334653
------------------------------------------------------------------------------

.

[Nick Cox]

Extract from the FAQ Advice:

6. Why are real names preferred?

You are asked to post on Statalist using your full real name, including given name(s) and a family name, such as "Ronald Fisher" or "Gertrude M. Cox". Giving full names is one of the ways in which we show respect for others and is a long tradition on Statalist. It is also much easier to get to know people when real names are used.

If you overlook this on first registration, it is easy to fix. Click on “Contact us” located at the bottom right-hand corner of every page.

[Felix Bittmann]

I would like to add that it would be interesting to know your total N and the group N. If the number is quite low it might be better to simplify the models and maybe go with a paired permutation test. This would allow a design that is a bit different but still interesting and probably more robust since the ANOVA has some parametric requirements I suppose.