Perform a paired ttest on two substracts of the same variable

[Luiza Lins]

I have a dataset that looks like this:

var1	var2	var3	attribute
1	0.93	0.88	1
1	0.76	0.20	1
1	0.40	0.18	0
0	0.34	0.91	1
0	0.09	0.51	0
...	...	...	...

For each variable (var1, var2, var3, etc), I need to perfom a ttest compairing its mean when attribute = 1 vs the mean of the whole variable. How can I do this?

[Ken Chui]

A couple points. The title says (bold is by me):

Perform a paired ttest on two substracts of the same variable

and then in the text, it says:

I need to perfom a ttest compairing its mean when attribute = 1 vs the mean of the whole variable. How can I do this?

I don't understand how this can be done in a paired t-test. Let's say var3 and var3|attribute == 1 is like the followings. The paired t-test result will be zero:

clear
input v1 v2 v3 attribute
1 0.93 0.88 1
1 0.76 0.20 1
1 0.40 0.18 0
0 0.34 0.91 1
0 0.09 0.51 0
end

keep v3 attribute
gen v3b = v3 if attribute == 1
list
ttest v3 == v3b

Results:

. list

     +----------------------+
     |  v3   attrib~e   v3b |
     |----------------------|
  1. | .88          1   .88 |
  2. |  .2          1    .2 |
  3. | .18          0     . |
  4. | .91          1   .91 |
  5. | .51          0     . |
     +----------------------+

. ttest v3 == v3b

Paired t test
------------------------------------------------------------------------------
Variable |     Obs        Mean    Std. err.   Std. dev.   [95% conf. interval]
---------+--------------------------------------------------------------------
      v3 |       3    .6633333    .2318285    .4015387   -.3341441    1.660811
     v3b |       3    .6633333    .2318285    .4015387   -.3341441    1.660811
---------+--------------------------------------------------------------------
    diff |       3           0           0           0           0           0
------------------------------------------------------------------------------
     mean(diff) = mean(v3 - v3b)                                  t =        .
 H0: mean(diff) = 0                              Degrees of freedom =        2

 Ha: mean(diff) < 0           Ha: mean(diff) != 0           Ha: mean(diff) > 0
 Pr(T < t) =      .         Pr(|T| > |t|) =      .          Pr(T > t) =      .

Did you mean independent sample t-test?

Also, please take some time to read the FAQ (http://www.statalist.org/forums/help). Pay attention to section 12 on using command dataex to provide Stata-readable data code. Like this:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(v1 v2 v3 attribute)
1 .93 .88 1
1 .76  .2 1
1  .4 .18 0
0 .34 .91 1
0 .09 .51 0
end

[Nick Cox]

clear
input v1 v2 v3 attribute
1 0.93 0.88 1
1 0.76 0.20 1
1 0.40 0.18 0
0 0.34 0.91 1
0 0.09 0.51 0
end

The only t test that I can imagine here with a clear rationale compares each other variable by attribute. I agree with Ken Chui that this kind of test isn't paired.

forval j = 1/3 {
    ttest v`j', by(attribute)
}

Cross-posted at https://stackoverflow.com/questions/...-same-variable

Please note our cross-posting policy, which is that you are asked to tell us about it. https://www.statalist.org/forums/help#crossposting