[R] Error message in `fitdistr` function in the MASS package

[kmmoon100]

I have one issue at the moment while I am using a function, fitdistr in
package MASS.
 I have 12 years of wind data with 30 min average interval.

My data look like below:

Time                windspeed

1/01/2005 0:00  15
1/01/2005 0:30  11
1/01/2005 1:00  11
1/01/2005 1:30  13
1/01/2005 2:00  15
1/01/2005 2:30  13
1/01/2005 3:00  13
1/01/2005 3:30  13
1/01/2005 4:00  15
1/01/2005 4:30  17
1/01/2005 5:00  17
1/01/2005 5:30  15
1/01/2005 6:00  13
1/01/2005 6:30  13
1/01/2005 7:00  15
1/01/2005 7:30  18
1/01/2005 8:00  18
1/01/2005 8:30  21
1/01/2005 9:00  21
1/01/2005 9:30  22
1/01/2005 10:00 24
1/01/2005 10:30 26
1/01/2005 11:00 24
1/01/2005 11:30 24
1/01/2005 12:00 24
1/01/2005 12:30 24
1/01/2005 13:00 17
1/01/2005 13:30 18
1/01/2005 14:00 21
1/01/2005 14:30 21
1/01/2005 15:00 22
1/01/2005 15:30 24
1/01/2005 16:00 24
1/01/2005 16:30 24
1/01/2005 17:00 21
1/01/2005 17:30 24
1/01/2005 18:00 22
1/01/2005 18:30 22
1/01/2005 19:00 21
1/01/2005 19:30 21
1/01/2005 20:00 15
1/01/2005 20:30 11
1/01/2005 21:00 13

My intention of using this function is to get daily Weibull parameters
(shape and scale) but it keeps giving me errors.

The code I put was:

walpeup_alldata <- read.csv("walpeup_excercise.csv", header=TRUE, sep=",")
walpeup_alldata$Date <- as.POSIXct(walpeup_alldata$Time, 
                                format = "%d/%m/%Y %H:%M", 
                                tz = "Australia/Melbourne")
walpeup_alldata$day_index <- paste(format(walpeup_alldata$Date, "%Y"), 
                                format(walpeup_alldata$Date, "%j"), sep =
"_")
walpeup_aggregation <- aggregate(walpeup_alldata$windspeed,
by=list(walpeup_alldata$day__index), FUN= function(x) {fitdistr(x,
densfun="weibull")} )

Interesting thing is when I used short duration of wind data, the code works
but produces lots of residuals. The residuals are like this.

33  2005_033    c(4.75967037103778, 36.0492072865698)
34  2005_034    c(7.71303917850763, 35.3491854177213)
35  2005_035    c(5.68407836144938, 28.721154798025)
36              c(0.485695856927515, 0.720482566441891)
37              c(0.479982752982148, 0.605561818452758)
38              c(0.398017326464339, 0.975977176044699)
39              c(0.603493688481038, 1.04186250740921)
40              c(0.787847162718371, 0.600969267921991)
71              c(0.235900465436553, 0.110092730139038, 0.110092730139038,
0.519095128546693)
72              c(0.230383443160322, 0.094083878067012, 0.094083878067012,
0.366705115967811)
73              c(0.15841779216582, 0.127050034063599, 0.127050034063599,
0.952531448160185)
106            -144.261676847584
107            -134.931368456724
108            -155.386471593028
109            -163.984332061118
172             48
173             48
174             48
175             48

However when I increase size of the data by more than few months, it starts
not to work again with providing below error message.

*Error in fitdistr(x, densfun = "weibull") : optimization failed
In addition: There were 50 or more warnings (use warnings() to see the first
50)*

I have taken out all zero values in advance. It worked well for another
dataset.
I don't know what the cause and the solution for the data are at the moment.

Thank you for your attention.



--
View this message in context: http://r.789695.n4.nabble.com/Error-message-in-fitdistr-function-in-the-MASS-package-tp4685862.html
Sent from the R help mailing list archive at Nabble.com.