Outliers

[jessisam]

How do I remove outliers form a group of data? I would like to remove the 2,5 % lowest and 2,5 % highest observations within this datagroup and make a new group out of this data without the outliers. I have tried this quote, but it doesn't work:

genr assets= @recode ( assets > @quantile ( assets, 0.025) or assets < 0.975)

Is there a quote to do this?

[startz]

How do I remove outliers form a group of data? I would like to remove the 2,5 % lowest and 2,5 % highest observations within this datagroup and make a new group out of this data without the outliers. I have tried this quote, but it doesn't work:

genr assets= @recode ( assets > @quantile ( assets, 0.025) or assets < 0.975)

Is there a quote to do this?

Code: Select all

smpl if assets > @quantile( assets, 0.025) or assets <@quantile( assets, 0.975)

[jessisam]

So there isn't a way to generate a new group without the outliers? The reason I ask is because I first want to remove the outliers and then divide the data set into 4 quantiles and see the descriptive statistics of this. Isn't there a quote like this that removes outliers:

genr assets = @recode (assets > @quantile (assets, .025) and assets < @quantile (assets, .975)

?

[EViews Gareth]

First off I believe there is a mistake in Startz's code - the OR should be an AND.

Secondly, yes you can create a series like that:
series test = @recode(assets > @quantile( assets, 0.025) and assets <@quantile( assets, 0.975), assets, na)


Thirdly, if all you want is descriptive statistics, you could use Startz's approach and set the sample, then calculate the statistics.

[diggetybo]

I have a follow up question,

How can we specify the sample to impose the quantile restrictions on all variables?

smpl if assets > @quantile( assets, 0.025) and assets <@quantile( assets, 0.975)

The above case is only for the series "assets"

Please advise,

Thanks

[EViews Gareth]

You'll have to just add more and statements to the sample.