How to do t-test for a subset of observation

[Alex Mai]

Dear statalists,

I have a question of how to run -ttest- for a subset of observations. I tried to add -if- after -ttest-, as the way in regression, but stata shows that -if- cannot follow -ttest-.

Thank you very much.

[Rich Goldstein]

you don't show your command something you know you should do; the following works fine:

sysuse auto, clear
ttest mpg if rep78<., by(foreign)

[Alex Mai]

you don't show your command something you know you should do; the following works fine:

sysuse auto, clear
ttest mpg if rep78<., by(foreign)

Dear Rich,

Thank you very much and sorry for the incomplete command in my post. Now I see that actually my mistake was to put -if- after -by()-.

[Marcos Almeida]

The "if qualifier" shall be typed before the comma.

Please see the example below:

. sysuse auto
(1978 Automobile Data)

. ttest mpg, by(foreign)

Two-sample t test with equal variances
------------------------------------------------------------------------------
   Group |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
Domestic |      52    19.82692     .657777    4.743297    18.50638    21.14747
 Foreign |      22    24.77273     1.40951    6.611187    21.84149    27.70396
---------+--------------------------------------------------------------------
combined |      74     21.2973    .6725511    5.785503     19.9569    22.63769
---------+--------------------------------------------------------------------
    diff |           -4.945804    1.362162               -7.661225   -2.230384
------------------------------------------------------------------------------
    diff = mean(Domestic) - mean(Foreign)                         t =  -3.6308
Ho: diff = 0                                     degrees of freedom =       72

    Ha: diff < 0                 Ha: diff != 0                 Ha: diff > 0
 Pr(T < t) = 0.0003         Pr(|T| > |t|) = 0.0005          Pr(T > t) = 0.9997

. ttest mpg if trunk<13, by(foreign)

Two-sample t test with equal variances
------------------------------------------------------------------------------
   Group |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
Domestic |      16     24.6875    1.116985    4.467941     22.3067     27.0683
 Foreign |      14    26.21429     1.34319    5.025758     23.3125    29.11607
---------+--------------------------------------------------------------------
combined |      30        25.4    .8611673    4.716808    23.63872    27.16128
---------+--------------------------------------------------------------------
    diff |           -1.526786    1.732872               -5.076412    2.022841
------------------------------------------------------------------------------
    diff = mean(Domestic) - mean(Foreign)                         t =  -0.8811
Ho: diff = 0                                     degrees of freedom =       28

    Ha: diff < 0                 Ha: diff != 0                 Ha: diff > 0
 Pr(T < t) = 0.1929         Pr(|T| > |t|) = 0.3858          Pr(T > t) = 0.8071

. ttest mpg, if trunk <13 by(foreign)
option if not allowed
r(198);

[Nick Cox]

That's not forbidden in itself:

ttest mpg , by(foreign), if rep78<.

is legal, but forget that I told you that.

[Marcos Almeida]

Nick Cox : wow, that is really impressive! Good to learn this. I checked putting the if qualifier in the middle, of no avail.

[Rich Goldstein]

generalizing Nick's note in #5, the rule in Stata is that options go after odd numbered commas and parts of the command are after even-numbered commas (counting 0 as even here); thus, one could take Nick's example and add ", une" and that would be fine also; as I have said previously, I do sometimes use this to advantage when trying to figure out exactly what I want - but my final do file will have only one comma (at most) per command

[Nick Cox]

As Rich implies, the comma is a switch between the main part of command and the options, and you can switch back and forth. I think this goes back to very early days when retyping a command was a bigger deal. Now it's best for style and clarity to do as Rich says, use at most one comma (of this kind).