can't detect outliers

#1
Hello, everyone,
I am doing an assignment where i need to check the data on the statistical assumptions,
I need to comment on the presence of outliers using histograms.
in the picture below, the ones circled red - are they all outliers, or only the top one?...:confused:
thanks!
 

Dason

Ambassador to the humans
#2
A bin in a histogram with a large frequency isn't indicative of the presence of outliers. The two circles at the tails of the distribution are more likely to be outliers but to be honest I wouldn't call anything in that histogram an outlier.
 
#5
Are you all using any particular definition of what an outlier is within your course?
well, it's a statistics course for English teachers, so we consider everything on quite a basic level.
so far we've regarded outliers as data which is "markedly different from the rest of the data" (this is from our book, Larsen-Hall).
 

noetsi

Fortran must die
#6
Its a lot easier to see outliers in Tukey box plot than a histogram. Hlsmith brings up an excellent point about the definitions. Boxplots for example define outliers based on how far a point is from the interquartile range (the later is the middle half of the data). It makes no distribution assumptions (like normality) about the data. Another common measure (if the standardized or studentized residuals are beyond either 2 or 3) does assume I believe normality as these are modified z scores essentially (on data that has been transformed various ways depending on if you use studentized or standardized values). Since they are residuals these measures are tied to where points lie relative to patterns in the data notably regression lines.

A histogram focuses not on outliers but the general distribution of the data. It has no specific mechanism to detect outliers (you can eye them but there is nothing formal like a boxplot or standardized residual rule).
 
#7
A histogram focuses not on outliers but the general distribution of the data. It has no specific mechanism to detect outliers (you can eye them but there is nothing formal like a boxplot or standardized residual rule).
thanks a lot, I have researched it and now I'm sure I will go for the boxplots :)
 

noetsi

Fortran must die
#8
There are many types of boxplots. Tukey's is probably the easiest to use (and certainly the be covered on line). It has a very simple formula for outliers. I use it constantly at work.