R help - Jun 2017 - Predict

Hi folks,

I have 25 stand height observations over 7 years period and daily
leafbiomass data during this period. I want to use the 25 plant height
observations as inputs and predict the daily stand height during the 7
years.


SH=matrix(data=NA , nrow = 2641, ncol = 1)
for (i in 1:2641) {
  SH<- predict(lm(height~Date, data=Stand_Height));

  dl=leafbiom$Date[i-1];
  de=leafbiom$Date[i];
SH[i]=sum(SH[leafbiom$Date==de&leafbiom$Date==dl])


}
SH


The SH output is the prediction of Stand height in 25 observations
only and provides NA for the remaining 2616 iterations.

Thanks for your help.



Ahmed Attia, Ph.D.
Agronomist & Soil Scientist

There are a bunch of things wrong here, although without a
reproducible example I can't really fix most of them.

- You're overwriting SH within the loop.
- You're running the regression 2641 times, even though the result
never changes.
- You're never predicting from your linear model using the other data
not in the regression.
- Leaf biomass data is never used for anything. I would have thought
that you would use leaf biomass as the predictor variable, not Date.
- I'm not sure why you want the cumulative sum of stand height; that
doesn't make sense to me.

I'm guessing you want:

height.model <- lm(height ~ leafbiomass, data = Stand_Height)
pred.height <- predict(height.model, leafbiom)

# not sure about the reasoning behind this
SH <- cumsum(pred.height)

You don't need a loop. Overwriting SH is the biggest R problem; the
rest of my questions have to do with what your objective actually is,
like what you are modeling and what you are doing with the
predictions. But my sample code might be enough to get you headed in
the right direction regardless.

Sarah

On Fri, Jun 30, 2017 at 9:27 AM, Ahmed Attia <ahmedatia80 at gmail.com>
wrote:
> Hi folks,
>
> I have 25 stand height observations over 7 years period and daily
> leafbiomass data during this period. I want to use the 25 plant height
> observations as inputs and predict the daily stand height during the 7
> years.
>
>
> SH=matrix(data=NA , nrow = 2641, ncol = 1)
> for (i in 1:2641) {
>   SH<- predict(lm(height~Date, data=Stand_Height));
>
>   dl=leafbiom$Date[i-1];
>   de=leafbiom$Date[i];
> SH[i]=sum(SH[leafbiom$Date==de&leafbiom$Date==dl])
>
>
> }
> SH
>
>
> The SH output is the prediction of Stand height in 25 observations
> only and provides NA for the remaining 2616 iterations.
>
> Thanks for your help.
>
>
>
> Ahmed Attia, Ph.D.
> Agronomist & Soil Scientist
>
-- 
Sarah Goslee
http://www.functionaldiversity.org

Sorry for the confusion, here is the edited question.

The data= Stand_Height (attached) is recorded from 12/1/2009 to
12/31/2015 (25 observations) and the other dataset (leafbiom) is
recorded from 10/7/2009 to 12/29/2016 (daily observations).

I want to use the 25 observations of stand height to predict the daily
stand height from 10/7/2009 to 12/29/2016. The daily stand height will
be multiplied by leaf biomass to produce a new variable.

I agree that a loop is not needed, would the forecast library help or
should I use predict library.

Stand_Height=ts(Stand_Height$height,start=2009,end = 2016,
                frequency =365)

plot(forecast(ets(Stand_Height),10))
a=seq(as.Date("2009-12-01"),by="weeks",length=11)
axis(1, at = a, labels = format(a, "%Y %b %d"), cex.axis=0.6)


#Error :$ operator is invalid for atomic vectors

Thanks


Ahmed Attia, Ph.D.
Agronomist & Soil Scientist






On Fri, Jun 30, 2017 at 10:37 AM, Sarah Goslee <sarah.goslee at gmail.com>
wrote:
> There are a bunch of things wrong here, although without a
> reproducible example I can't really fix most of them.
>
> - You're overwriting SH within the loop.
> - You're running the regression 2641 times, even though the result
> never changes.
> - You're never predicting from your linear model using the other data
> not in the regression.
> - Leaf biomass data is never used for anything. I would have thought
> that you would use leaf biomass as the predictor variable, not Date.
> - I'm not sure why you want the cumulative sum of stand height; that
> doesn't make sense to me.
>
> I'm guessing you want:
>
> height.model <- lm(height ~ leafbiomass, data = Stand_Height)
> pred.height <- predict(height.model, leafbiom)
>
> # not sure about the reasoning behind this
> SH <- cumsum(pred.height)
>
> You don't need a loop. Overwriting SH is the biggest R problem; the
> rest of my questions have to do with what your objective actually is,
> like what you are modeling and what you are doing with the
> predictions. But my sample code might be enough to get you headed in
> the right direction regardless.
>
> Sarah
>
> On Fri, Jun 30, 2017 at 9:27 AM, Ahmed Attia <ahmedatia80 at
gmail.com> wrote:
>> Hi folks,
>>
>> I have 25 stand height observations over 7 years period and daily
>> leafbiomass data during this period. I want to use the 25 plant height
>> observations as inputs and predict the daily stand height during the 7
>> years.
>>
>>
>> SH=matrix(data=NA , nrow = 2641, ncol = 1)
>> for (i in 1:2641) {
>>   SH<- predict(lm(height~Date, data=Stand_Height));
>>
>>   dl=leafbiom$Date[i-1];
>>   de=leafbiom$Date[i];
>> SH[i]=sum(SH[leafbiom$Date==de&leafbiom$Date==dl])
>>
>>
>> }
>> SH
>>
>>
>> The SH output is the prediction of Stand height in 25 observations
>> only and provides NA for the remaining 2616 iterations.
>>
>> Thanks for your help.
>>
>>
>>
>> Ahmed Attia, Ph.D.
>> Agronomist & Soil Scientist
>>
> --
> Sarah Goslee
> http://www.functionaldiversity.org