basic truncated normal dist #36
Conversation
phobson
force-pushed
the
add-truncnorm
branch
3 times, most recently
from
f0343da to
69569ce
Compare
| if lower is not None and mu is not None and sigma is not None: | ||
| a = (lower - mu) / sigma | ||
|
|
||
| if upper is not None and mu is not None and sigma is not None: |
There was a problem hiding this comment.
@ev-br sorry to bother you. any thoughts on how to deal with missing parameters when fitting? this method seems almost useless if you need 3 parameters just to fix one.
In other words, scipy.stats.truncnorm.fit(fa=<guess>) will work, but paramnormal.truncated_normal.fit(lower=<guess>) won't since you also need mu and sigma to estimate a.
On the otherhand, if you're fitting truncnorm distributions and you have a guess for a, does it follow that you implicitly have a guess for sigma and mu?
There was a problem hiding this comment.
Geez, that's confusing! A quick thought, likely rubbish --- how about say, fixing upper and lower from the extent of the data and then calling truncnorm.fit_loc_scale to estimate mu and sigma?
| References | ||
| ---------- | ||
| http://scipy.github.io/devdocs/generated/scipy.stats.truncnorm | ||
| https://en.wikipedia.org/wiki/Rice_distribution |
There was a problem hiding this comment.
Good catch and I didn't know about fit_loc_scale! Will check it out.
On Fri, Jul 29, 2016 at 5:03 AM Evgeni Burovski [email protected]
wrote:
In paramnormal/dist.py
#36 (comment):
- and
- .. math::
b = \frac{x_{\mathrm{upper bound}} - \mu}{\sigma}
- Since
aandbare directly linked to the location and scale- of the distribution as well as the lower and upper limits,
- respectively, it's difficult to use the
fitmethod of this- distirbution without either knowing a lot about it
a priorior- assuming just as much.
- References
Rice?
—
You are receiving this because you were assigned.
Reply to this email directly, view it on GitHub
https://github.com/phobson/paramnormal/pull/36/files/c30f4bcfcc2a0e8b426c322d7eb383556547d7af#r72780884,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABHCowITbToKZ4dE3_fR0uKTVadumbB6ks5qaev4gaJpZM4HvL4U
.
closes #35