1
$\begingroup$

I have monthly precipitation for one site over a 35 year period. I would like to test if the total precipitation for a specific year is equal to the mean total precipitation for the rest of the period of record. In addition, I would also like to split the data into wet season [Nov-April] vs dry season [May-Oct].

I realize there are likely issues to take into account regarding auto-correlation, but I am yet analyze these types of data personally.

Would someone please be kind enough to point me in the right direction?

Thanks ahead of time.

EDIT: IrishStat, The data is below, I use R, so below was created using dput(). I could figure this out, but it would take me a while and I would feel more confident if I were shown once.

PPT <- structure(list(YEAR = 1965:2010, JAN = c(3.15, 4.81, 4.94, 2.49, 4.26, 4.31, 3.96, 5.06, 3, 3.58, 6.27, 1.79, 6.14, 14.43, 5.04, 6.22, 0.94, 3.23, 6.6, 5.74, 5.08, 3.8, 6.35, 4.38, 0.87, 8.66, 15.48, 13.08, 7.52, 5.02, 5.45, 5.23, 5.88, 16.33, 5.18, 2.95, 3.33, 4.91, 2.49, 5.75, 2.08, 2.62, 4.38, 7.02, 2.16, 6.73), FEB = c(5.95, 9.37, 6.09, 2.84, 3.87, 4.27, 6.72, 3.78, 7.09, 6.56, 3.97, 2.39, 1.87, 4.48, 7.73, 1.46, 11.59, 9.58, 11.61, 5.95, 7.03, 4.66, 8.96, 6.4, 1.64, 10.3, 2.4, 8.06, 4.29, 1.88, 2.84, 6.36, 6.52, 6.19, 0.82, 0.87, 3.56, 2.65, 7.03, 8.12, 5.6, 4.22, 2.07, 4.88, 4.86, 5.43), MAR = c(4.89, 2.58, 0.69, 1.98, 8.94, 7.17, 4.9, 7.12, 8.23, 6.88, 6.83, 9.95, 4.73, 5.79, 9.29, 11.6, 2.59, 7.84, 9.38, 5.58, 5.06, 3.91, 5.43, 6.36, 4.32, 10.41, 4.89, 2.9, 7.31, 7.83, 10.03, 10.2, 3.75, 6.11, 5.79, 4.12, 8.73, 4.18, 5.2, 0.68, 4.28, 0.39, 0.54, 4.32, 14.43, 5.16), APR = c(0.61, 2.45, 1.68, 2.25, 4.55, 1.94, 0.48, 2.29, 9.22, 5.08, 6.84, 1.96, 1.83, 4.76, 5.77, 13.53, 1.24, 2.9, 12.97, 3.19, 1.01, 3.33, 0.63, 3.71, 2.88, 2.47, 9.1, 2.72, 2.55, 4.74, 6.71, 11.68, 6.32, 4.54, 0.11, 1.22, 0.3, 3.12, 3.06, 2.32, 20.5, 6.09, 3.43, 5.49, 2.07, 2.04), MAY = c(0.42, 5.66, 2.86, 2.7, 9.89, 2.86, 4.53, 7.8, 2.63, 3.4, 11.11, 10.7, 2.8, 10.71, 5.5, 9.61, 11.88, 1.64, 1.54, 2.49, 7.85, 5.35, 12.35, 0.31, 7, 4.94, 13.75, 2.4, 5.81, 2.89, 6.04, 0.48, 7.96, 0.81, 3.15, 0.65, 0.55, 2.34, 5.84, 2.03, 7.08, 3.44, 1.86, 9.26, 7.3, 7.03), JUN = c(15.43, 2.34, 4.33, 3.79, 4.27, 9.77, 2.82, 4.42, 6.91, 3.01, 3.52, 4.19, 0.71, 12.32, 3.37, 3.81, 2.51, 7.53, 6.6, 2.8, 8.08, 5.54, 15.58, 5.91, 18.52, 6.22, 5.85, 4.62, 3.38, 7.32, 4.53, 7.42, 5.88, 2.23, 8.32, 4.21, 13.77, 3.4, 9.54, 10.8, 10.44, 1.33, 6.35, 3.32, 3.69, 5.04), JUL = c(13.72, 7.6, 3.69, 6.95, 10.88, 1.93, 6.88, 6.56, 7.67, 5.29, 10.03, 8.03, 5.04, 23.67, 8.4, 4.06, 5.05, 11.94, 0.82, 3.9, 8, 4.29, 5.16, 4.8, 8.92, 5.84, 8.64, 5.03, 8.08, 10.94, 8.65, 6.55, 28.58, 6.17, 9.71, 3.21, 13.69, 10.76, 18.44, 4.73, 11.43, 5.34, 7.09, 5.4, 5.18, 2.18), AUG = c(10.1, 8.51, 7.88, 3.57, 12.91, 11.84, 6.4, 4.14, 4.3, 6.75, 8.69, 2.08, 9.33, 6.24, 5.41, 1.31, 6.28, 6.03, 6.08, 14.23, 5.78, 5.01, 12.48, 13.19, 2.17, 2, 6.74, 6.2, 7.57, 6.49, 9.61, 5.99, 1.25, 5.9, 5.99, 3.02, 11.02, 5.83, 5.2, 8.28, 11.4, 7.13, 5.96, 14.14, 6.24, 10.33), SEP = c(10.74, 3.43, 6.37, 4.19, 5.52, 5.53, 11.07, 1.01, 10.13, 7.6, 12.88, 6.48, 7.53, 3.97, 9.45, 5.28, 1.11, 2.31, 7.02, 0.57, 6.01, 3.07, 2.4, 16.12, 4.49, 1.64, 3.13, 1.44, 4.99, 1.46, 3.05, 7.72, 1.29, 24.11, 2.27, 9.53, 4.66, 14.77, 3.66, 12.63, 4.7, 5.25, 6.64, 7.67, 8.16, 6.33), OCT = c(2.57, 3.84, 7, 0.38, 1.75, 7.1, 0, 2.77, 1.28, 0.56, 5.52, 6.81, 4.13, 0, 0.69, 3.77, 1.9, 2.22, 5.53, 1.78, 13.08, 5.89, 0.23, 1.97, 2.28, 2.88, 2.28, 3.85, 5.3, 5.88, 12.13, 2.93, 4.88, 1.72, 4.42, 0.39, 3.43, 8.44, 2.68, 2.44, 0.25, 3.76, 8.78, 4.14, 7.67, 1.68), NOV = c(0.89, 4.88, 0.58, 3.83, 0.8, 2.69, 2.29, 4.64, 3.59, 3.05, 8.83, 5.57, 6.73, 3.76, 7.55, 5.43, 0.81, 4.78, 6.84, 2.26, 2.98, 6.06, 5.07, 4.44, 7.76, 1.84, 3.8, 11.7, 1.8, 3.71, 11.63, 2.13, 12.03, 6.27, 1.02, 7.34, 1.78, 5.69, 3.84, 11.64, 3.22, 3.7, 3.91, 3.47, 3.82, 4.73), DEC = c(3.87, 4.92, 5.77, 6.2, 8.27, 5.21, 5.93, 7.51, 7.07, 3.58, 3.63, 4.95, 4.76, 3.93, 1.96, 1.67, 5.88, 8.04, 5.5, 1.7, 3.29, 4.62, 5.3, 2.47, 5.21, 2.37, 2.87, 5.28, 4.73, 2.05, 4.84, 6.14, 2.52, 5.35, 5.4, 3.49, 2.74, 4.87, 4.13, 5.58, 3.9, 4.51, 5.77, 3.34, 13.52, 2.72)), .Names = c("YEAR", "JAN", "FEB", "MAR", "APR", "MAY", "JUN", "JUL", "AUG", "SEP", "OCT", "NOV", "DEC"), class = "data.frame", row.names = c(NA, -46L))
$\endgroup$
2
  • $\begingroup$ If you want to add more info, please edit your question (note you can always do it regardless of your reputation). $\endgroup$ – user88 Aug 31 '11 at 14:39
  • $\begingroup$ OK, sorry about the data format and if I did something incorrect with regards to me editing the original post. I have emailed you a CSV file. $\endgroup$ – Patrick Aug 31 '11 at 20:40
2
$\begingroup$

Patrick, You could add a series 0,0,0,0,0,....1,1,1,1,1,1,1,1,1,1,1,1, ,0,0,0,0, where the 1's reflect the year you are trying to test. This variable would be the X variable in an ARMAX model and it's significance would test the hypothesis of that year being exceptional. Be careful to include both ARMA structure and any Level Shifts / Trends / Pulses and Seasonal PUlses that might also be necessary. If you don't have the software/smarts to be able to do this post the data to the web and I will look at it and report back and perhaps suggest some ways for you to do this yourself.

There is no reason to split the data as a forecast for May may depend on April. A comprehensive analysis will allow you to test myths.

$\endgroup$
3
  • $\begingroup$ :chi It appears that you deleted my post as being inappropriate. Please let me know at dave@autobox.com why you did this so as I will not make the "mistake" again . Thanks $\endgroup$ – IrishStat Aug 31 '11 at 12:46
  • $\begingroup$ :chi . Now I see ..It looks like a nearly identical answer ! Thanks for pulling it. $\endgroup$ – IrishStat Aug 31 '11 at 12:58
  • $\begingroup$ Thanks for the offer IrishStat; I accept if you can find the time (I added the data above). I am really interested in whether the period of time from March 2009 to June 2010 is extreme (wetter or drier) relative to the rest of the period of record. I really appreciate this; I will use your suggestions as a starting point for my self-education on the topic. $\endgroup$ – Patrick Aug 31 '11 at 14:34
0
$\begingroup$

This is a second answer because there is not enough room in a "comment" . The data is monthly ( starting 1965/1) http://www.autobox.com/patrick/patrick.xls . After a brief review of the data it was apparent that there was no strong seasonal structure ! Oftentime there may be particular/individual months of the year that may have repetitive patterns thus we added 11 dummies to test the JAN-NOV effects. Additionally the OP had wanted to test the hypothesis that the period between 2009/3 and 2010/6 had "been statistically unusual". We show a test of that hypthesis ( t=-1.58 )at enter image description here .suggesting non-significance. We found two months of the year to have statistically significant effects: (March (1.02) and October (-1.04)) . Additionally there were a number of months that had one-time effects (Pulses) and three Level Shifts in the series earmarking points in time that the mean of the series had shifted.. these three points in time were

   1975/12   -1.4
   1985/4    +2.1
   1999/7    -1.9

The final plot of residuals visually suggests randomness :enter image description here

Finally the acf of the final model residuals supported the hypothesis if randomness enter image description here . The very small acf values appear to be statistically significant but this is an artifact of the large sample size (n=552) this the standard error of the acf is approximately 1/sqrt(552) or about .04.

The bottom line , Patrick , is there is no assignable cause to the special period and that there are two unusual months of the year and there are essentially 4 regimes ( 3 level shifts ) besides some very unusual and unpredictable values/readings using the history of the series months. Precipitation might depend on something else besides it's history !.

I think that this data set should be scrutinized by readers of the list. Such scrutiny can only confirm/test the analysis conducted here. Different readers have access to different tools and sometimes these lead to different results/findings.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.