COMPGROUPS.NET | Search | Post Question | Groups | Stream | About | Register

### Multiple time-series

• Follow

```HI,
I'm new to matlab and I would like to analyse multiple concurrent time-series (e.g wind speed, air pressure & visibility).
I suspect tstool is the way to go, but I do not know how to import multiple data sets in the same matrix.
Also, the time intervals on the datasets differ - 5 minute intervals for met data and 3 minute interval for the visibilitydata. Can matlab interpolate?
If there is a step-by-step guide anywhere, please feel free to point me to it!
```
 0
Reply Sanette 8/11/2010 8:50:08 AM

```"Sanette " <blugekko11@gmail.com> wrote in message <i3to80\$rfa\$1@fred.mathworks.com>...
> HI,
> I'm new to matlab and I would like to analyse multiple concurrent time-series (e.g wind speed, air pressure & visibility).
> I suspect tstool is the way to go, but I do not know how to import multiple data sets in the same matrix.
> Also, the time intervals on the datasets differ - 5 minute intervals for met data and 3 minute interval for the visibilitydata. Can matlab interpolate?
> If there is a step-by-step guide anywhere, please feel free to point me to it!

Lets see, How do you want to analyse the differente data sets? one by one (ARIMA models or linear regressions) or you are working with VAR models for time series? This is my first question.The fact that the measurements are in different point intervals affects in different ways depending on your model at hand. So can you be more specific about your model and what do you want to do.
```
 0
Reply Rogelio 8/11/2010 10:48:05 AM

```> Lets see, How do you want to analyse the differente data sets? one by one (ARIMA models or linear regressions) or you are working with VAR models for time series? This is my first question.The fact that the measurements are in different point intervals affects in different ways depending on your model at hand. So can you be more specific about your model and what do you want to do.

Thank you for the response! At the end of the day I'd like to correlate all the meteorology time series (1 to 1 at first) with the visibility time series. I'm trying to see whether specific good visibility events are related to specific meteorology variables. I'm not looking to predict any trends; but rather to establish the correlation between visibility & e.g. wind speed
```
 0
Reply Sanette 8/11/2010 12:04:08 PM

```"Sanette " <blugekko11@gmail.com> wrote in message <i3u3jo\$pcs\$1@fred.mathworks.com>...
>
> > Lets see, How do you want to analyse the differente data sets? one by one (ARIMA models or linear regressions) or you are working with VAR models for time series? This is my first question.The fact that the measurements are in different point intervals affects in different ways depending on your model at hand. So can you be more specific about your model and what do you want to do.
>
> Thank you for the response! At the end of the day I'd like to correlate all the meteorology time series (1 to 1 at first) with the visibility time series. I'm trying to see whether specific good visibility events are related to specific meteorology variables. I'm not looking to predict any trends; but rather to establish the correlation between visibility & e.g. wind speed

Ok, Im not an expert in meteorology but i can help u with the modelliing issues. The "easy" model you can use is a OLS, i.e. linear regression model. Where you have a dependent variable ( I presume that in your case is visibility) and two independent variables ( in this case will be air preassure and wind speed). So, this model will "tell you" if the wind and the preasure have any influence in the visibility and the degree of it. It might be the case that a linear model, such as OLS, does not fit your data, i.e. the model is not helpful for your purposes. This you will know it by conducting the proper diagnostic tests for the OLS. I suggest u to get and introductory book to statistics and read a bit about OLS or wikipedia, so you can interprete correctly your results. Of course, this might be not what you are looking. Let me know if this sounds like what you need.
```
 0
Reply Rogelio 8/11/2010 3:46:04 PM

```> Ok, Im not an expert in meteorology but i can help u with the modelliing issues. The "easy" model you can use is a OLS, i.e. linear regression model. Where you have a dependent variable ( I presume that in your case is visibility) and two independent variables ( in this case will be air preassure and wind speed). So, this model will "tell you" if the wind and the preasure have any influence in the visibility and the degree of it. It might be the case that a linear model, such as OLS, does not fit your data, i.e. the model is not helpful for your purposes. This you will know it by conducting the proper diagnostic tests for the OLS. I suggest u to get and introductory book to statistics and read a bit about OLS or wikipedia, so you can interprete correctly your results. Of course, this might be not what you are looking. Let me know if this sounds like what you need.<

Thank you, I'll read up on it! But it definitely sounds like the first step in modelling I'd look into. Now just for the data incorporation.
Thank you kindly!
```
 0
Reply Sanette 8/12/2010 9:29:20 AM

```"Sanette " <blugekko11@gmail.com> wrote in message <i40etg\$jcd\$1@fred.mathworks.com>...
>
> > Ok, Im not an expert in meteorology but i can help u with the modelliing issues. The "easy" model you can use is a OLS, i.e. linear regression model. Where you have a dependent variable ( I presume that in your case is visibility) and two independent variables ( in this case will be air preassure and wind speed). So, this model will "tell you" if the wind and the preasure have any influence in the visibility and the degree of it. It might be the case that a linear model, such as OLS, does not fit your data, i.e. the model is not helpful for your purposes. This you will know it by conducting the proper diagnostic tests for the OLS. I suggest u to get and introductory book to statistics and read a bit about OLS or wikipedia, so you can interprete correctly your results. Of course, this might be not what you are looking. Let me know if this sounds like what you need.<
>
> Thank you, I'll read up on it! But it definitely sounds like the first step in modelling I'd look into. Now just for the data incorporation.
> Thank you kindly!

how do you have your data? which format? you mention that the frequencies are not equal, right? which frequencies do you have? As I mendtion I do not know anything about meteorology but I pressume that it is inapropriate to use different frequencies in a OLS. The results will be not meaningful. Do you have any book of applied statistics in meteorology?
```
 0
Reply Rogelio 8/12/2010 1:18:07 PM

5 Replies
239 Views

(page loaded in 0.036 seconds)

Similiar Articles:

7/8/2012 8:47:25 PM

 Reply: