Confidence Intervals for t-test
-------------------------------

         ^tci^ continuous_var group_var [^if^ exp] [^in^ range]

^tci^ gives 99.9%, 99%, 95%, 90%, 80%, 65% and 50% confidence intervals, and
t-test results, ^assuming^ that the variances are equal.  A graph showing
these limits is produced at the end; this is saved as #.gph, where # is
replaced by 1...n for the number of continuous variables you run this ado for
file, ^without^ leaving STATA.  If you leave STATA and then re-enter and try
to run this you will get an error message when the program tries to save the
first graph unless you have changed the name of the graphs previously saved.
The ado file is written this way assuming you want to keep old graphs, or at
least you do not want to automatically destroy them.

Your group_var will automatically be, temporarily, replaced by a 0-1 version
so that the differences shown in the output will be correct.  The 0-1 code
will be 0 for the smallest of the first two codes found and the 1 will be for
the largest, even if you have more than two groups in your original group_var.
If you want this handled some other way, then make your own dummy variable
prior to calling ^tci^.  Your data set is replaced as it was saved on disk at
the end of the run.

Note that the xlabels and the top labels are not complete since this would
lead to overlap of .99 and .999--however there are tick marks at each of the
confidence levels graphed; seven levels are graphed, the same as the numbers
that are shown (see the examples below).

Examples using auto.dta (without the graphs):

. us auto
(1978 Automobile Data)

. tci mpg for

                                          Lower Limit     Upper Limit
The level(.999) confidence interval is:      0.2701         9.6215
The level(.99) confidence interval is:       1.3417         8.5499
The level(.95) confidence interval is:       2.2304         7.6612
The level(.90) confidence interval is:       2.6760         7.2156
The level(.80) confidence interval is:       3.1840         6.7077
The level(.65) confidence interval is:       3.6644         6.2272
The level(.50) confidence interval is:       4.0224         5.8692

mean for first group is   19.8269    mean for second group is   24.7727

The difference is:    4.9458   t:   3.6308  and p-value is 0.0005 (N=74)

(1978 Automobile Data)

The second example uses ^if^ to test two of five groups:

. tci mpg rep78 if rep78==3 | rep78==5

Following are confidence intervals for the difference:
                                          Lower Limit     Upper Limit
The level(.999) confidence interval is:     0.7051         15.1555
The level(.99) confidence interval is:      2.5028         13.3578
The level(.95) confidence interval is:      3.8780         11.9826
The level(.90) confidence interval is:      4.5548         11.3058
The level(.80) confidence interval is:      5.3186         10.5420
The level(.65) confidence interval is:      6.0352          9.8254
The level(.50) confidence interval is:      6.5663          9.2943

mean for first group is   19.4333    mean for second group is   27.3636

The difference is:   7.9303   t:  3.9584  and p-value is 0.0004 (N=41)

(1978 Automobile Data)

The data set name is shown at the end since the ado file re-sets the data file
to the way it was before running the ado file; thus, all new variables
disappear.  However, the nonlocal macros, including the graph counter, do not
disappear; they can be examined by typing "mac li" at the STATA command.