Author: Ajay Ohri

http://about.me/ajayohri

ROC Curve

ROC Curve is a nice modeling concept to know as it will used practically in nearly all models

irrespective of spoefic technique and irrespective of statistical software.

We use the Wikipedia for referring to easy to implement statistics rather than crusty

thick books which seem prohibitely dense and opaque to outsiders

-This is how you define the ROC Curve.

actual value
p n total
prediction
outcome		 p’ True
Positive		 False
Positive		 P’
n’ False
Negative		 True
Negative		 N’
total P N

true positive (TP)

eqv. with hit
true negative (TN)
eqv. with correct rejection
false positive (FP)
eqv. with false alarm, Type I error
false negative (FN)
eqv. with miss, Type II error
true positive rate (TPR)
eqv. with hit rate, recall, sensitivity
TPR = TP / P = TP / (TP + FN)
false positive rate (FPR)
eqv. with false alarm rate, fall-out
FPR = FP / N = FP / (FP + TN)
accuracy (ACC)
ACC = (TP + TN) / (P + N)
specificity (SPC)
SPC = TN / (FP + TN) = 1 ? FPR
positive predictive value (PPV)
eqv. with precision
PPV = TP / (TP + FP)

Here is a good java enabled page to calculate the ROC Curve.

http://www.rad.jhmi.edu/jeng/javarad/roc/JROCFITi.html

And in case any one asks, ROC stands for Receiver Operating Characteristic. ……

Facebook Design Changes

Cooler. Sleeker. Check out www.new.facebook.com

Here’s my new profile– Its like it just re designed by some Apple guy (what  is Apple doing that they are keeping secret…anything at all in social networking)

Mark Zuckenberg will have that 15 billion Dollar valuation yet, even if he has to slay all the Open Social Networks put together. No wonder MS and FB are coming closer…enemy of enemy of my friend and all that.

Besides the fact that founders of Microsoft and Facebook are Harvard droputs.

That’s my new business plan. Get into Harvard, be the founder of a new company, drop out.

Statistically it should work !Or get in Stanford and drop out .

A Prayer to the Killed-Updated

A Prayer to the Killed
Kill the mullahs who rain terror
Kill the Jews who fire the missiles
Inspiring the mullahs to hit back
Kill the white men who arm them both
Kill the brown, the black, and the yellow men
Who are so busy killing,
They forget the art of living
Kill all the killers
Kill em all
BUT after you hunt down the hunters, my friend
Say a prayer for the innocent gone
The ones who didn”t know why they were dying
Who jumped a hundred floors
To avoid the burning of their skin.
Say a prayer for the people who were trying to save them
And couldn”t save themselves
A prayer for the people who struggled and died in the air
Rather than kill some more on the ground
Say a prayer for the dead, the dying and the coming dead
Kill the killers
But know what forced them to hate you so much
And pray for the killed
Whether they are yours or the enemy”s, my friend
Amen.

 

(written after September 11,2001. Unfortunately it is relevant again due to bomb blasts in India)

One More Cool Widget

Another Cool and By Now Quite Familiar Widget (example bloglines etc)

Note this is an example only. Decision Stats site is non commercial.
But if you slog and blog, earn some more peanuts here.

Enjoy!!

More fun than plain ol Adsense ..say what ?

Segmenting Models : When and Why

Creating segmented models in SAS is quite easy, with a by group processing . It is less easy in other softwares , but that is understandable given that

the first generic rule of segmentation is

1) each segment has statistically similar characteristics .

2) different segments have statistically different characteristics .

This means that just using Proc freq to
check response rate versus
independent variable is not a good way
to check the level of difference.
Proc univariate with plot option and
a by group processing
is actually a better way to test out
because it is a combination of means ,
median analysis 
(measures of central value) but also
box plot ,normal distributions and standard deviations
(measures of dispersion).

Proc freq with cross tab is incredibly powerful to decide whether to create a model in the first place. But fine tuning of decisions on segments is better done with proc univariate. The SAS equivalent for clustering of course remains Proc Fastclus and family which will be dealt in a separate post.

(Note :lovely image that explains the above from Dr Ariel Shamir’s home page (he is a research expert on Visual Succinct Representation of Information   ————-from Israel, land of the brave and intelligent).

A Picture is truly worth a thousand words (or posts !).)

If America had a Queen and other stuff

If America had a Queen, and a King.

—–Would Price Charles be called Prince Charlie. Would Prince William be called Prince Bill.

If God could do anything

—–Could he make stone so heavy that he couldnot lift it himself.

If the Devil convinced everyone he didnot exist

—–Would Satan and Santa be spelled the same.

If you had to live your life again

—–Would you spend less time on the internet and more time with your family.

If the English copyrighted the English language just like people copyright anything they create-

Would we pay 1 cent per word, or switch to Chinese (because Chinese would be cheaper).

If you are getting very bored

—–Would you read the next post. Or wouldnt you ?

Iwannacrib has just been revamped for the third time at a strategic level.

Watch this space….or subscribe from the form in the footer.

Ajay

Chief Imagination Officer.