P value for trend

[Dearbhla Kelly]

I would be very grateful if someone could help me with the code for this - I'm reporting the incidence of stroke in patients with diabetes versus no diabetes in 5 year bands, and would like to calculate the P value for trend over time if possible. I've tried the ptrend command but I'm confused about the other syntax. My data looks something like this:

Year Strokes in non-diabetics Strokes in diabetics
2002-2007 238/72243 343/4030
2007-2012 336/74945 282/4375
2012-2017 431/76741 169/4809

Thanks much in advance!

[Nick Cox]

The intervals are ambiguous. Can you get annual data?

[Joseph Coveney]

I . . . would like to calculate the P value for trend over time if possible.

Take a look at the linear component of the orthogonal polynomial contrasts below.


.ÿ
.ÿversionÿ16.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿinputÿstr9ÿepochÿlong(count01ÿcount0tÿcount11ÿcount1t)

ÿÿÿÿÿÿÿÿÿepochÿÿÿÿÿÿÿcount01ÿÿÿÿÿÿÿcount0tÿÿÿÿÿÿÿcount11ÿÿÿÿÿÿÿcount1t
ÿÿ1.ÿ"2002-2007"ÿ238ÿ72243ÿ343ÿ4030
ÿÿ2.ÿ"2007-2012"ÿ336ÿ74945ÿ282ÿ4375
ÿÿ3.ÿ"2012-2017"ÿ431ÿ76741ÿ169ÿ4809
ÿÿ4.ÿend

.ÿ
.ÿencodeÿepoch,ÿgenerate(epo)ÿlabel(Epochs)

.ÿforvaluesÿiÿ=ÿ0/1ÿ{
ÿÿ2.ÿÿÿÿÿÿÿÿÿgenerateÿlongÿcount`i'0ÿ=ÿcount`i'tÿ-ÿcount`i'1
ÿÿ3.ÿÿÿÿÿÿÿÿÿdropÿcount`i't
ÿÿ4.ÿ}

.ÿ
.ÿquietlyÿreshapeÿlongÿcount0ÿcount1,ÿi(epo)ÿj(stk)

.ÿquietlyÿreshapeÿlongÿcount,ÿi(epoÿstk)ÿj(dbs)

.ÿ
.ÿlogitÿstkÿi.epo##i.dbsÿ[fweight=count],ÿnolog

LogisticÿregressionÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿ237,143
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿLRÿchi2(5)ÿÿÿÿÿÿÿÿ=ÿÿÿÿ2419.51
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿÿÿÿÿÿÿ=ÿÿÿÿÿ0.0000
Logÿlikelihoodÿ=ÿ-9364.0993ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿPseudoÿR2ÿÿÿÿÿÿÿÿÿ=ÿÿÿÿÿ0.1144

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿstkÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿepoÿ|
ÿÿ2007-2012ÿÿ|ÿÿÿ.3093149ÿÿÿ.0848832ÿÿÿÿÿ3.64ÿÿÿ0.000ÿÿÿÿÿ.1429468ÿÿÿÿÿ.475683
ÿÿ2012-2017ÿÿ|ÿÿÿÿ.535769ÿÿÿÿ.080925ÿÿÿÿÿ6.62ÿÿÿ0.000ÿÿÿÿÿ.3771589ÿÿÿÿ.6943791
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿ1.dbsÿ|ÿÿÿ3.337382ÿÿÿ.0860363ÿÿÿÿ38.79ÿÿÿ0.000ÿÿÿÿÿ3.168754ÿÿÿÿÿ3.50601
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿepo#dbsÿ|
2007-2012#1ÿÿ|ÿÿ-.6096034ÿÿÿ.1190893ÿÿÿÿ-5.12ÿÿÿ0.000ÿÿÿÿ-.8430141ÿÿÿ-.3761927
2012-2017#1ÿÿ|ÿÿ-1.473502ÿÿÿ.1259691ÿÿÿ-11.70ÿÿÿ0.000ÿÿÿÿ-1.720397ÿÿÿ-1.226607
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿ_consÿ|ÿÿÿ-5.71222ÿÿÿ.0649274ÿÿÿ-87.98ÿÿÿ0.000ÿÿÿÿ-5.839476ÿÿÿ-5.584965
------------------------------------------------------------------------------

.ÿ
.ÿ//ÿInÿtheÿoriginalÿestimationÿmetric
.ÿcontrastÿp.epo#dbs

Contrastsÿofÿmarginalÿlinearÿpredictions

Marginsÿÿÿÿÿÿ:ÿasbalanced

--------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿÿchi2ÿÿÿÿÿP>chi2
---------------------+----------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿepo#dbsÿ|
ÿÿÿ(linear)ÿ(joint)ÿÿ|ÿÿÿÿÿÿÿÿÿÿ1ÿÿÿÿÿÿ136.83ÿÿÿÿÿ0.0000
(quadratic)ÿ(joint)ÿÿ|ÿÿÿÿÿÿÿÿÿÿ1ÿÿÿÿÿÿÿÿ1.50ÿÿÿÿÿ0.2200
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿJointÿÿ|ÿÿÿÿÿÿÿÿÿÿ2ÿÿÿÿÿÿ137.39ÿÿÿÿÿ0.0000
--------------------------------------------------------

.ÿ
.ÿ//ÿInÿtermsÿofÿproportions
.ÿquietlyÿmarginsÿepo#dbs,ÿpost

.ÿcontrastÿp.epo#dbs

ContrastsÿofÿadjustedÿpredictionsÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿ237,143
ModelÿVCEÿÿÿÿ:ÿOIM

Expressionÿÿÿ:ÿPr(stk),ÿpredict()

--------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿÿchi2ÿÿÿÿÿP>chi2
---------------------+----------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿepo#dbsÿ|
ÿÿÿ(linear)ÿ(joint)ÿÿ|ÿÿÿÿÿÿÿÿÿÿ1ÿÿÿÿÿÿ103.22ÿÿÿÿÿ0.0000
(quadratic)ÿ(joint)ÿÿ|ÿÿÿÿÿÿÿÿÿÿ1ÿÿÿÿÿÿÿÿ0.90ÿÿÿÿÿ0.3416
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿJointÿÿ|ÿÿÿÿÿÿÿÿÿÿ2ÿÿÿÿÿÿ117.31ÿÿÿÿÿ0.0000
--------------------------------------------------------

.ÿ
.ÿexit

endÿofÿdo-file


.

Nick is doubly right, by the way: not only do your intervals overlap, but it's not clear how the data came about (complex survey?) or whether a logistic model (e.g., binomial likelihood) is even appropriate.