[Tutor] Replacing a value in a list

[nzbz xx]

Assuming that when there are consecutive missing values, they should be
replaced with adjacent valid values e.g [1,2,-999,-999,5] should give
[1,2,2,5,5]. And given [1,-999,-999,-999,5], the middle missing value would
take the average of index 1 & 3. This should get an output of [1, 2 , 2,
3.5, 5]. How should it be coded for it to solve from the outer elements
first to the inner elements?

On Fri, Aug 13, 2021 at 8:52 PM Mats Wichmann <mats at wichmann.us> wrote:

> On 8/13/21 1:14 AM, nzbz xx wrote:
> > I came across this problem: Given that -999 is a missing value in a
> > dataset, replace this value with the mean of its adjacent values. For
> > example [1, 10, -999, 4, 5] should yield 7.
> >
> > This is what i got so far. The problem is that i don't know what to do
> when
> > there are consecutive -999 in the list e.g.  [1, 10, -999, -999, -999]
>
> For the problem as stated, you probably want to raise an error. But
> since the problem statement doesn't say, there isn't really an "answer"
> we can give.  It's not actually a Python problem.
>
> This is data science... in the real world, collected data may have
> missing points, and what you do about  missing points affects your
> conclusions, and the choices are almost certainly colored by an
> understanding of the data. I think that lecture (which I'm not
> qualified to give anyway) probably belongs in an entirely different
> forum.  One of the approaches is to "impute" a value for a missing
> element, and one way to do that is to use a mean of nearest neighbors as
> in your problem. Even when you've chosen that specific approach there
> are variances: the size of window used for nearest-neighbor may smooth
> too much if too large or too little if too small. If the neighbors are
> missing too, you would probably conclude that missing-neighbors was not
> a good choice for imputation.  Maybe there's a different algorithm to
> try? And just maybe, that the dataset might be of such low quality that
> you can't draw any reasonable conclusions from it
>
>
> _______________________________________________
> Tutor maillist  -  Tutor at python.org
> To unsubscribe or change subscription options:
> https://mail.python.org/mailman/listinfo/tutor
>