Interpreting Autocorrelation Correlogram

#1
I am new to time series and need to interpret the following graph: http://imgur.com/AaZjTnY

I know the graph tells me something about the presence of autocorrelation, but I'm not sure how to interpret it. Also, I know it tells me how many lags would be optimal to include, but -once again- I have no idea how to interpret my findings.

Thank
 

Miner

TS Contributor
#2
If I am reading your graph correctly, you do not have any autocorrelation in your time series. Autocorrelation, if present, would appear in Lag 1 and progress for n lags then disappear. Normally, the graph would have limits. If the bar at a particular lag exceeded the limit, it would indicate the presence of autocorrelation. If you had three lags showing autocorrelation, you would include three lags in your model.
 

noetsi

Fortran must die
#3
It is more typical for the ACF or PACF to run horizontally than vertically:p There should be band on both sides of the ACF and PACF showing the 2 SE region - these are critical for analysis and not shown here. The PACF is more valuable for determing the specific order of the AR.

I thought only first order AR was signficant at lag 1 not all orders of AR generally. Also note that seasonality is a separate issue.
 

Miner

TS Contributor
#4
Attached is an example of an autocorrelation graph using Minitab. This example shows statistically significant autocorrelation at 2 lags, but not at 3 lags.