Colleagues, Thanks all for the responses. I am monitoring the daily total number of defects per sample unit. I need to know whether this daily defect proportion is trending upward (a bad thing for a manufacturing process). My first thought was to use either a u or a u' control chart for this. As far as I know, u or u' charts are poor to detect drifts. This is why I chose to use prop.trend.test to detect trends in proportions. While prop.trend.test can confirm the existence of a trend, as far as I know, it is left to the user to determine what direction that trend is. One way to illustrate trending is of course to plot the data and use geom_smooth and method lm For the non-statisticians in my group, I've found that using this method along with the p-value of prop.trend.test, makes it easier for the users to determine the existence of trending and its direction. If there are any other ways to do this, please let me know. Thomas Subia On Thursday, September 7, 2023 at 10:31:27 AM PDT, Rui Barradas <ruipbarradas at sapo.pt> wrote: ?s 14:23 de 07/09/2023, Thomas Subia via R-help escreveu:> > Colleagues > >? ?Consider > smokers? <- c( 83, 90, 129, 70 ) > patients <- c( 86, 93, 136, 82 ) > >? ?prop.trend.test(smokers, patients) > >? ?Output: > >? ???? Chi-squared Test for Trend inProportions > >? ?data:??smokers out of patients , > > using scores: 1 2 3 4 > > X-squared = 8.2249, df = 1, p-value = 0.004132 > >? ?# trend test for proportions indicates proportions aretrending. > >? ?How does one identify the direction of trending? >? ?# prop.test indicates that the proportions are unequal but doeslittle to indicate trend direction. > All the best, > Thomas Subia > > > ??? [[alternative HTML version deleted]] > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code.Hello, By visual inspection it seems that there is a decreasing trend. Note that the sample estimates of prop.test and smokers/patients are equal. smokers? <- c( 83, 90, 129, 70 ) patients <- c( 86, 93, 136, 82 ) prop.test(smokers, patients)$estimate #>? ? prop 1? ? prop 2? ? prop 3? ? prop 4 #> 0.9651163 0.9677419 0.9485294 0.8536585 smokers/patients #> [1] 0.9651163 0.9677419 0.9485294 0.8536585 plot(smokers/patients, type = "b") Hope this helps, Rui Barradas
You might want to consider exponential smoothing models such as Holt's (Double Exponential Smoothing). This method continually updates the trend parameter, and you can monitor the most recent value (for sign, or magnitude, or both). In R, some choices to fit the Holt model: 1. stats::HoltWinters() and set the gamma argument to FALSE for a non-seasonal model 2. forecast::holt() HTH, Eric On Fri, Sep 8, 2023 at 8:23?AM Thomas Subia via R-help <r-help at r-project.org> wrote:> > Colleagues, > > Thanks all for the responses. > > I am monitoring the daily total number of defects per sample unit. > I need to know whether this daily defect proportion is trending upward (a bad thing for a manufacturing process). > > My first thought was to use either a u or a u' control chart for this. > As far as I know, u or u' charts are poor to detect drifts. > > This is why I chose to use prop.trend.test to detect trends in proportions. > > While prop.trend.test can confirm the existence of a trend, as far as I know, it is left to the user > to determine what direction that trend is. > > One way to illustrate trending is of course to plot the data and use geom_smooth and method lm > For the non-statisticians in my group, I've found that using this method along with the p-value of prop.trend.test, makes it easier for the users to determine the existence of trending and its direction. > > If there are any other ways to do this, please let me know. > > Thomas Subia > > > > > > > > > > > > > On Thursday, September 7, 2023 at 10:31:27 AM PDT, Rui Barradas <ruipbarradas at sapo.pt> wrote: > > > > > > ?s 14:23 de 07/09/2023, Thomas Subia via R-help escreveu: > > > > Colleagues > > > > Consider > > smokers <- c( 83, 90, 129, 70 ) > > patients <- c( 86, 93, 136, 82 ) > > > > prop.trend.test(smokers, patients) > > > > Output: > > > > Chi-squared Test for Trend inProportions > > > > data: smokers out of patients , > > > > using scores: 1 2 3 4 > > > > X-squared = 8.2249, df = 1, p-value = 0.004132 > > > > # trend test for proportions indicates proportions aretrending. > > > > How does one identify the direction of trending? > > # prop.test indicates that the proportions are unequal but doeslittle to indicate trend direction. > > All the best, > > Thomas Subia > > > > > > [[alternative HTML version deleted]] > > > > ______________________________________________ > > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > > https://stat.ethz.ch/mailman/listinfo/r-help > > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > > and provide commented, minimal, self-contained, reproducible code. > Hello, > > By visual inspection it seems that there is a decreasing trend. > Note that the sample estimates of prop.test and smokers/patients are equal. > > > smokers <- c( 83, 90, 129, 70 ) > patients <- c( 86, 93, 136, 82 ) > > prop.test(smokers, patients)$estimate > #> prop 1 prop 2 prop 3 prop 4 > #> 0.9651163 0.9677419 0.9485294 0.8536585 > > smokers/patients > > #> [1] 0.9651163 0.9677419 0.9485294 0.8536585 > > plot(smokers/patients, type = "b") > > > > Hope this helps, > > Rui Barradas > > ______________________________________________ > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code.
Yes, this was written a bit bone-headed (as I am allowed to say...)
If you look at the code, you will see inside:
a <- anova(lm(freq ~ score, data = list(freq = x/n, score =
as.vector(score)),
weights = w))
and the lm() inside should give you the direction via the sign of the regression
coefficient on "score".
So, at least for now, you could just doctor a copy of the code for your own
purposes, as in
fit <- lm(freq ~ score, data = list(freq = x/n, score = as.vector(score)),
weights = w)
a <- anova(fit)
and arrange to return coef(fit)["score"] at the end. Something like
structure(... estimate=c(lpm.slope=coef(fit)["score"]) ....)
(I expect that you might also extract the t-statistic from coef(summary(fit))
and find that it is the signed square root of the Chi-square, but I won't
have time to test that just now.)
-pd
> On 8 Sep 2023, at 07:22 , Thomas Subia via R-help <r-help at
r-project.org> wrote:
>
> Colleagues,
>
> Thanks all for the responses.
>
> I am monitoring the daily total number of defects per sample unit.
> I need to know whether this daily defect proportion is trending upward (a
bad thing for a manufacturing process).
>
> My first thought was to use either a u or a u' control chart for this.
> As far as I know, u or u' charts are poor to detect drifts.
>
> This is why I chose to use prop.trend.test to detect trends in proportions.
>
> While prop.trend.test can confirm the existence of a trend, as far as I
know, it is left to the user
> to determine what direction that trend is.
>
> One way to illustrate trending is of course to plot the data and use
geom_smooth and method lm
> For the non-statisticians in my group, I've found that using this
method along with the p-value of prop.trend.test, makes it easier for the users
to determine the existence of trending and its direction.
>
> If there are any other ways to do this, please let me know.
>
> Thomas Subia
>
>
>
>
>
>
>
>
>
>
>
>
> On Thursday, September 7, 2023 at 10:31:27 AM PDT, Rui Barradas
<ruipbarradas at sapo.pt> wrote:
>
>
>
>
>
> ?s 14:23 de 07/09/2023, Thomas Subia via R-help escreveu:
>>
>> Colleagues
>>
>> Consider
>> smokers <- c( 83, 90, 129, 70 )
>> patients <- c( 86, 93, 136, 82 )
>>
>> prop.trend.test(smokers, patients)
>>
>> Output:
>>
>> Chi-squared Test for Trend inProportions
>>
>> data: smokers out of patients ,
>>
>> using scores: 1 2 3 4
>>
>> X-squared = 8.2249, df = 1, p-value = 0.004132
>>
>> # trend test for proportions indicates proportions aretrending.
>>
>> How does one identify the direction of trending?
>> # prop.test indicates that the proportions are unequal but
doeslittle to indicate trend direction.
>> All the best,
>> Thomas Subia
>>
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
> Hello,
>
> By visual inspection it seems that there is a decreasing trend.
> Note that the sample estimates of prop.test and smokers/patients are equal.
>
>
> smokers <- c( 83, 90, 129, 70 )
> patients <- c( 86, 93, 136, 82 )
>
> prop.test(smokers, patients)$estimate
> #> prop 1 prop 2 prop 3 prop 4
> #> 0.9651163 0.9677419 0.9485294 0.8536585
>
> smokers/patients
>
> #> [1] 0.9651163 0.9677419 0.9485294 0.8536585
>
> plot(smokers/patients, type = "b")
>
>
>
> Hope this helps,
>
> Rui Barradas
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes at cbs.dk Priv: PDalgd at gmail.com