.-
help for ^tgraph^                                               (Patrick Roysto
> n)
.-

Plot graph with transformed axes --------------------------------

^tgraph^ yvarlist xvar [^if^ exp] [^in^ range]^,^ [^xtr^ans^(^x_transf^)^ ^x > la^bel^(^xlablist^)^ ^ytr^ans^(^y_transf^)^ ^yla^bel^(^ylablist^)^ graph_options ]

Description -----------

^tgraph^ produces a graph with one or two non-standard axes, whose labels are o > n the same scale as yvarlist and/or xvar.

Options -------

^xtrans()^ defines the transformation (function of xvar) to be used. The argume > nt of the function is represented by ^@@^. Standard Stata syntax is used. If y > ou specify ^xtrans()^ then you must also specify ^xlabel()^.

^ytrans()^ defines the transformation (function of all the variables in yvarlis > t) to be used. The rules for defining the function are the same as for ^xtrans()^. If you specify ^ytrans()^ then you must also specify ^ylabel()^ > .

^xlabel()^, ^ylabel()^ are the required axis labels.

graph_options are any of Stata's @graph, twoway@ options except for ^xlabel()^ > and ^ylabel()^.

Remarks -------

Stata provides automatic axis scaling only for linear (natural) and logarithmic scales, the latter via the ^xlog^, ^ylog^ and ^rlog^ options of ^graph^. You ma > y want more. For example, a square root transformation of the response or of a covar- iate is sometimes useful, and you would like an axis to be scaled accordingly but with the labels on the original scale of measurement. ^tgraph^ allows you t > o do this in a flexible way. It exploits Stata's valuable facility to label an axis automatically according to the score labels for the variable in question.

For example, suppose you want to transform a variable called ^weeks^ to a quad- ratic polynomial function of it, for which the coefficients of the linear and quadratic terms are stored in global macros called ^$b1^ and ^$b2^. You then wa > nt to plot a variable called ^femur^ against ^weeks^ on the quadratically transfor > med scale, labelling values on the X-axis with ^10,20,30,40^ weeks.

The quadratic function of ^weeks^ may be represented by the following expressio > n:

^$b1*weeks + $b2*weeks^^2^

The plot is produced using ^tgraph^ as follows:

. ^tgraph femur weeks, xlabel(10(10)40) xtrans($b1*@@ + $b2*@@^^2)^

Examples --------

. ^tgraph femur weeks, xlabel(10,20,30,40) xtrans(sqrt(@@))^ . ^tgraph femur weeks, labels(10,20,30,40) xtrans($b1*@@ + $b2*@@^^2)^ . ^tgraph y1 y2 y3 x, connect(lll) ylabel(1,2,3,4) ytrans(-1/@@)^ ^xlabel(10(10)40) xtrans(sqrt(@@))^

Also see --------

Manual: [U] graph On-line: help for @label@, @graph@