Error when fitting the scipy.stats.erlang distribution

The name of the pictureThe name of the pictureThe name of the pictureClash Royale CLAN TAG#URR8PPP

Error when fitting the scipy.stats.erlang distribution



I'm new to this, but I have some data that I am trying to fit to various distributions.



Right now I'm just trying to figure out what I am doing wrong. Here is an example.


data = np.array([-0.00363649, 0.00329399, 0.00648754, -0.00115784,
-0.00020446, -0.0223319])

scipy.stats.erlang.fit(data)



This returns the warning:


RuntimeWarning: The shape parameter of the erlang distribution has been given a non-integer value 2.1.
RuntimeWarning: The shape parameter of the erlang distribution has been given a non-integer value 1.9.
.
.
.



It does this a lot of times starting with 2.1 and then descending. I'm a little confused about what is happening.




1 Answer
1



By definition, the Erlang distribution has an integer shape parameter. Other than that, it is the same the gamma distribution.



The SciPy implementation of the Erlang distribution is basically the same as the gamma distribution, but it generates a warning when you give a non-integer value for the shape parameter. The fit method of the SciPy distributions does not do discrete optimization, so it will pass non-integer values to the Erlang distribution methods, and that results in the warnings.


fit



To avoid this, use the gamma distribution instead of the Erlang distribution.






By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Comments

Post a Comment

Popular posts from this blog

Executable numpy error

Image
Clash Royale CLAN TAG #URR8PPP Executable numpy error I'm trying to create an executable file for my code. I've already tried with cx_Freeze and with pyinstaller. Both gives me the same error, which is: "Missing required dependencies 0.format(missing_dependencies))" PS C:UsersGustavoDesktopbuildexe.win32-3.6> python AgendaOficial.py C:UsersGustavoAppDataLocalProgramsPythonPython36-32python.exe: can't open file 'AgendaOficial.py': [Errno 2] No such file or directory PS C:UsersGustavoDesktopbuildexe.win32-3.6> .AgendaOficial.exe Traceback (most recent call last): File "C:UsersGustavoAppDataLocalProgramsPythonPython36-32libsite-packagescx_Freezeinitscripts__startup__.py", line 14, in run module.run() File "C:UsersGustavoAppDataLocalProgramsPythonPython36-32libsite-packagescx_FreezeinitscriptsConsole.py", line 26, in run exec(code, m. dict ) File "AgendaOficial.py", line 8, in File "C:UsersGustavoAppDataLocal
Read more

PySpark count values by condition

Image
Clash Royale CLAN TAG #URR8PPP PySpark count values by condition I have a DataFrame, a snippet here: [['u1', 1], ['u2', 0]] basically one string ('f') and either a 1 or a 0 for second element ('is_fav'). What I need to do is grouping on the first field and counting the occurrences of 1s and 0s. I was hoping to do something like num_fav = count((col("is_fav") == 1)).alias("num_fav") num_nonfav = count((col("is_fav") == 0)).alias("num_nonfav") df.groupBy("f").agg(num_fav, num_nonfav) It does not work properly, I get in both cases the same result which amounts to the count for the items in the group, so the filter (whether it is a 1 or a 0) seems to be ignored. Does this depend on how count works? count 1 Answer 1 There is no filter here. Both col("is_fav") == 1 and col("is_fav") == 0) are just boolean expressions and count doesn't really care about their value as long a
Read more

Trying to Print Gridster Items to PDF without overlapping contents

Image
Clash Royale CLAN TAG #URR8PPP Trying to Print Gridster Items to PDF without overlapping contents I want to print gridster items to pdf using chome browser. Unfortunately the item overlaps on next page. Is there any way to make gridster item go to next page if it doesnt fit in the current page when printing to pdf? Screenshot By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.
Read more