Time series regression in R: help from scratch

I am very inexperienced in R, but I am told that this is the main statistical package. I studied it for the sake of research, but I'm stumped about time series data, regression in particular. I know how to do multiple regression, and I somewhat know how to make predictions using SARIMA models, but I'm not sure how to do multiple regression with time series.

Here is an example of my data. I always import from CSV.

     HomRate   Unemp    Av_Schl     GNI_perCapita   
2000    5.5    4.099    12.7        36930   
2001    6.6    4.800    N/A         37860   
2002    5.6    5.900    N/A         38590   
2003    5.6    6.099    N/A         39960   
2004    5.5    5.599    12.87391    42260   
2005    5.6    5.199    12.8        44740   
2006    5.8    4.699    12.96034    47390
2007    5.6    4.699    N/A         48420
2008    5.4    5.900    13.20913    48640
2009    5.0    9.399    13.29049    47250

      

Homicide, or HomRate, is the dependent variable and the others are independent. For the sake of this, let's say there are no transformations in the data.

From my limited understanding, instead of using lm

as in a few linear regressions, I use tslm

from a package forecast

. However, my data is not considered time series data according to R; How can i do this? None of the examples I've found actually show me the underlying data, so I don't know what the corresponding ts compatible data looks like.

If I get the command tslm

from the ground, will the rest of the logic be like multiple regression? (i.e. model=tslm(HomRate~Unemp+Av_Schl+...)

) Or is it very different in terms of encoding?

Thank you very much and please let me know if you need more details.

+3


source to share


1 answer


I am currently learning R and my main focus is time series analysis and I come across a lot of packet conflicts with dates and ggplot2.

As of Nov-2017, it looks like the least-risk approach is outlined in the R Times Series Study Guide by Matthew Mal.

Basic steps:



  • Data import
  • Load it into an object xts

    to view and filter it.
  • Converting an object xts

    to a standard R objectts

  • This object is ts

    more suitable for Rob Hyndman R packages and for use in ggplot2

Please give feedback if there is a better, lower risk way for people new to R.

0


source







All Articles