ISYEl 6501l Midterml Examl 2l (Latestl 2025/l 2026l Update)l Introductionl tol Analyticsl Modelingl Reviewl |Questionsl &l Answers|l Gradel A|l 100%l Correctl (Verifiedl Solutions)
Q:l Whatl arel thel prosl andl consl ofl LASSOl andl elasticl net
Answer:
Theyl arel slowerl butl helpl makel modelsl thatl makel betterl predictions
Q:l Whichl twol methodsl doesl elasticl netl lookl likel itl combinesl andl whatl arel thel downsidesl froml it?
Answer:
Ridgel Regressionl andl LASSO.
Advantages:l variablel selectionl froml LASSOl andl Predictivel benefitsl ofl LASSO.
Disadvantages:l Arbitrarilyl rulesl outl somel correlatedl variablesl likel LASSOl (don'tl knowl whichl onel thatl isl leftl outl shouldl be);l Underestimatesl coefficientsl ofl veryl predictivel variablesl likel Ridgel Regresison
Q:l Whatl arel somel downsidesl ofl surveys?
Answer:
Evenl ifl youl whatl appearsl tol bel al representativel samplel inl simplel ways,l maybel itl isn'tl inl morel complexl ways.
Q:l Ifl we'rel testingl tol seel whetherl redl carsl selll forl higherl pricesl thanl bluel cars,l wel needl tol accountl forl thel typel andl agel ofl thel carsl inl ourl datal set.l Thisl isl called:
- / 3
Answer:
Controlling
Q:l whatl isl al blockingl factor
Answer:
al sourcel ofl variabilityl thatl isl notl ofl primaryl interestl tol thel experimenter
Q:l whatl isl anl examplel ofl al blockingl factor
Answer:
Thel typel ofl car,l sportsl carl orl familyl car,l isl al blockingl factorl thatl itl couldl accountl forl somel ofl thel differencel betweenl redl carsl andl bluel cars.l Becausel sportsl carsl arel morel likelyl tol bel red;l ifl wel accountl forl thel difference,l wel canl reducel thel variabilityl inl ourl estimates
Q:l Underl whatl conditionsl shouldl youl runl A/Bl tests
Answer:
Whenl youl canl collectl datal quickly.l Whenl thel datal isl representativel andl thel amountl ofl datal isl smalll comparedl tol thel wholel population
Q:l Dol youl havel tol decidel thel samplel sizel aheadl ofl timel forl A/Bl tests
Answer:
no,l andl wel canl runl thel hypothesisl testl anytimel wel want
Q:l Whatl isl fulll factoriall design
Answer:
youl testl everyl combinationl andl thenl usel ANOVAl tol determinel importancel ofl eachl factor
Q:l Whatl isl fractionall factoriall design 2 / 3
Answer:
whenl youl testl al subsetl ofl thel entirel setl ofl combinations
Q:l Whatl isl al balancedl design?
Answer:
Youl testl eachl choicel thel samel #l ofl timesl andl eachl pairl ofl choicesl thel samel #l ofl times
Q:l Whenl isl regressionl effectivel workl welll tol determinel importantl factors?
Answer:
Ifl therel aren'tl significantl interactionsl betweenl thel factors.
Q:l whatl isl exploration?
Answer:
focusingl onl gettingl morel information;l inl thisl case,l tol determinel withl morel certaintyl whichl adl isl reallyl thel best
Q:l whatl isl exploitation
Answer:
we'rel focusedl onl gettingl immediatel value;l inl thisl example,l tol showl thel addl thatl seemsl tol bel doingl bestl sol far,l becausel itl seemsl tol bel mostl likelyl tol bel clicked.
Q:l whatl isl thel multi-armedl banditl approachl andl howl doesl itl balancel explorationl andl exploitation.
Answer:
Wel startl withl nol infol andl havel anl equall probabilityl ofl selectingl eachl alternative.l Afterl performingl somel tests,l we'vel gottenl morel information,l sol wel canl updatel thel probabilitiesl ofl eachl onel beingl bestl andl startl assigningl newl testsl accordingl tol thosel probabilities.l Wel keepl testingl multiplel alternatives;l so,l we'rel stilll doingl exploration.l Butl wel makel itl morel likelyl tol pickl thel bestl onesl sol we'rel alsol doingl exploitation
- / 3