Lecture 4

Question 1

There are many ways to communicate the same thing

Answer 1

like 75
can be 75
or seventy five
or IIIIIIIIIIIIIII(75 sticks)

Question 2

mark

Answer 2

basic geometric element that depict items or links

Question 3

channel

Answer 3

Control the appearance of marks

Question 4

Mark types in table datasets

Answer 4

mark represent an item

Question 5

Mark types in network datasets

Answer 5

represent either an item (node) or a link that represents a relationship between items

Question 6

A connection mark shows a pairwise relationship between two items using a line

Answer 6

true

Question 7

A containment mark shows hierarchical relationships using areas (nested connection marks within each other at multiple levels)

Answer 7

True

Question 8

Bootstrap presents a way to resample your sample over and overhowever it does it with replacement unlike cross validation

Answer 8

true

Question 9

If n is way larger than predictors, your model will perform well

Answer 9

True

Question 10

If I want to reduce the number of predictors I can use subset selections, one of these is

Answer 10

True

Question 11

We have 3 ways for subset selection

Answer 11

1- Best subset selection
2- Forward subset selection
3- Backward subset selection

Question 12

Best Subset Regression (selection):

Answer 12

Minimize RSS ( square distance between what we predict and the values that we have in the data)

Maximize adjRsquare( how much proportion the predictors explain the variance in the response)

Question 13

For best subset selection we need leaps

Answer 13

TRUE

Question 14

adjusted r-square

Answer 14

Adjusting R-square by the number of predictors

Question 15

Lasso approach

Answer 15

It will shrink coefficients and it will do feature selection

Question 16

Models generated from Lasso are sparse

Answer 16

Sparse meaning less number of predictors

Question 17

In both ridge and lasso we want to minimize RSS

Answer 17

Lasso: absolute value
Ridge: square

Question 18

In case not all predictors relate to the response Lasso will prerform better

Answer 18

True

Question 19

How to select Lambda

Answer 19

we do vlalidation and choose one of least error