{"id":34958,"date":"2019-10-31T22:01:31","date_gmt":"2019-10-31T19:01:31","guid":{"rendered":"https:\/\/prohoster.info\/blog\/kogda-stoit-proveryat-gipotezu-o-ne-menshej-effektivnosti\/"},"modified":"2019-10-31T22:01:31","modified_gmt":"2019-10-31T19:01:31","slug":"kogda-stoit-proveryat-gipotezu-o-ne-menshej-effektivnosti","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/kogda-stoit-proveryat-gipotezu-o-ne-menshej-effektivnosti","title":{"rendered":"Kur \u00ebsht\u00eb e udh\u00ebs t\u00eb verifikohet hipoteza p\u00ebr efikasitetin e nj\u00ebjt\u00eb?","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"Kur
\nArtikulli nga ekipi i Stitch Fix sugjeron p\u00ebrdorimin e qasjes s\u00eb studimeve klinike me efektivitet t\u00eb ngjash\u00ebm (provat e jo-inferioritetit) n\u00eb testet A\/B p\u00ebr marketing dhe produkte. Kjo qasje \u00ebsht\u00eb me t\u00eb v\u00ebrtet\u00eb e aplikueshme kur po testojm\u00eb nj\u00eb zgjidhje t\u00eb re q\u00eb ka avantazhe, t\u00eb paqart\u00ebsuara nga testet.<\/p>\n

Shembulli m\u00eb i thjesht\u00eb \u00ebsht\u00eb reduktimi i kostove. P\u00ebr shembull, ne automatizojm\u00eb procesin e caktimit t\u00eb m\u00ebsimit t\u00eb par\u00eb, por nuk duam t\u00eb ndajm\u00eb shum\u00eb konvertimin e p\u00ebrgjithsh\u00ebm. Ose testojm\u00eb ndryshime q\u00eb orientohen n\u00eb nj\u00eb segment p\u00ebrdoruesish, duke kaluar koh\u00ebn e monitorimit q\u00eb konvertimet n\u00eb segmentet e tjera mos u ulen shum\u00eb (kur testojm\u00eb disa hipoteza, mos harrojm\u00eb rregullat p\u00ebrpjes\u00ebtuese).<\/p>\n

Zgjedhja e kufirit t\u00eb duhur t\u00eb jo-inferioritetit shton v\u00ebshtir\u00ebsi t\u00eb tjera n\u00eb faz\u00ebn e projektimit t\u00eb testit. Pyetja, si t\u00eb zgjedhim \u0394, n\u00eb artikull nuk \u00ebsht\u00eb shum\u00eb mir\u00eb e shpjeguar. Duket se ky zgjedhje nuk \u00ebsht\u00eb plot\u00ebsisht e qart\u00eb as n\u00eb provat klinike. P\u00ebrmbledhje<\/a><\/noindex> publikimeve mjek\u00ebsore mbi non-inferiority tregon se vet\u00ebm n\u00eb gjysm\u00ebn e publikimeve zgjedhja e kufirit arsyetohet, dhe shpesh k\u00ebto arsyetimet jan\u00eb t\u00eb paqarta ose jo t\u00eb detajuara.<\/p>\n

Megjithat\u00eb, ky qasje duket interesante, pasi duke e zvog\u00ebluar madh\u00ebsin\u00eb e nevojshme t\u00eb mostr\u00ebs, mund t\u00eb rris\u00eb shpejt\u00ebsin\u00eb e testimit dhe, k\u00ebshtu, shpejt\u00ebsin\u00eb e marrjes s\u00eb vendimeve. \u2014 Darya Mukhina, analiste e produkteve n\u00eb aplikacionin mobil Skyeng.<\/i>
\n<\/a><\/noindex>
\nEkipi i Stitch Fix p\u00eblqen t\u00eb testoj\u00eb gj\u00ebra t\u00eb ndryshme. E gjith\u00eb bashk\u00ebsia teknologjike n\u00eb parim p\u00eblqen t\u00eb kryej\u00eb teste. Cila version i faqes attrakton m\u00eb shum\u00eb p\u00ebrdorues \u2013 A apo B? A sjell versioni A i modelit t\u00eb rekomandimit m\u00eb shum\u00eb par\u00e0 se versioni B? Gati gjithmon\u00eb p\u00ebr t\u00eb verifikuar hipotezat ne p\u00ebrdorim qasjen m\u00eb t\u00eb thjesht\u00eb nga kursi bazik i statistik\u00ebs: <\/p>\n

\"Kur
\n
\nMegjithat\u00eb, ne rrall\u00ebher\u00eb e p\u00ebrdorim k\u00ebt\u00eb term, dhe ky lloj testimi quhet \u00abtestimi i hipotez\u00ebs s\u00eb superioritetit\u00bb. Me k\u00ebt\u00eb qasje, ne supozojm\u00eb se nuk ka asnj\u00eb diferenc\u00eb midis dy varianteve. Ne mbajm\u00eb k\u00ebt\u00eb ide dhe e refuzojm\u00eb vet\u00ebm n\u00ebse t\u00eb dh\u00ebnat e marra rezultojn\u00eb t\u00eb mjaftueshme p\u00ebr k\u00ebt\u00eb \u2013 dometh\u00ebn\u00eb, tregojn\u00eb se nj\u00eb nga variantet (A ose B) \u00ebsht\u00eb m\u00eb e mir\u00eb se tjetra.<\/p>\n

Kontrollimi i hipotez\u00ebs p\u00ebr supremacin\u00eb \u00ebsht\u00eb i p\u00ebrshtatsh\u00ebm p\u00ebr zgjidhjen e nj\u00eb s\u00ebr\u00eb problemesh. Ne l\u00ebshojm\u00eb versionin B t\u00eb modelit rekomandues vet\u00ebm n\u00ebse \u00ebsht\u00eb duksh\u00ebm m\u00eb i mir\u00eb se versioni A q\u00eb po p\u00ebrdorim. Por n\u00eb disa raste, ky qasje nuk funksionon aq mir\u00eb. Le t\u00eb shqyrtojm\u00eb disa shembuj.<\/p>\n

1) Ne p\u00ebrdorim nj\u00eb sh\u00ebrbim t\u00eb jasht\u00ebm<\/b>, i cili ndihmon n\u00eb identifikimin e kartelave bankare t\u00eb falsifikuara. Ne gjet\u00ebm nj\u00eb sh\u00ebrbim tjet\u00ebr q\u00eb kushton ndjesh\u00ebm m\u00eb pak. N\u00ebse sh\u00ebrbimi m\u00eb i lir\u00eb funksionon po aq mir\u00eb sa ai q\u00eb p\u00ebrdorim tani, ne do ta zgjedhim at\u00eb. Ai nuk duhet patjet\u00ebr t\u00eb jet\u00eb m\u00eb i mir\u00eb se sh\u00ebrbimi q\u00eb po p\u00ebrdorim. <\/p>\n

2) Ne duam t\u00eb heqim dor\u00eb nga burimi i t\u00eb dh\u00ebnave<\/b> A dhe t\u00eb z\u00ebvend\u00ebsojm\u00eb at\u00eb me burimin e t\u00eb dh\u00ebnave B. Ne mund t\u00eb vonojm\u00eb heqjen e dor\u00eb nga A, n\u00ebse B jep rezultate shum\u00eb t\u00eb dob\u00ebta, por t\u00eb vazhdojm\u00eb t\u00eb p\u00ebrdorim A nuk \u00ebsht\u00eb e mundur. <\/p>\n

3) Ne do t\u00eb donim t\u00eb kalonim nga qasja n\u00eb modelimin<\/b>A n\u00eb qasjen B, jo sepse presim rezultate m\u00eb t\u00eb mira nga B, por sepse kjo na jep m\u00eb shum\u00eb fleksibilitet operativ. Nuk kemi asnj\u00eb arsye p\u00ebr t\u00eb supozuar se B do t\u00eb jet\u00eb m\u00eb keq, por nuk do t\u00eb b\u00ebjm\u00eb kalimin n\u00ebse do t\u00eb jet\u00eb k\u00ebshtu.<\/p>\n

4) Ne kemi b\u00ebr\u00eb disa p\u00ebrmir\u00ebsime cil\u00ebsore<\/b> n\u00eb dizajnin e faqes s\u00eb internetit (versioni B) dhe besojm\u00eb se ky version e kap\u00ebrcen versionin A. Nuk presim ndryshime n\u00eb konvertim apo ndonj\u00eb tregues t\u00eb r\u00ebnd\u00ebsish\u00ebm t\u00eb performanc\u00ebs, p\u00ebr t\u00eb cilat zakonisht vler\u00ebsojm\u00eb faqen e internetit. Por besojm\u00eb se ka p\u00ebrfitime n\u00eb parametrat q\u00eb jan\u00eb ose t\u00eb pakapshme, ose teknologjit\u00eb tona nuk jan\u00eb mjaftuesh\u00ebm t\u00eb avancuara p\u00ebr t'i matur ato.<\/p>\n

N\u00eb k\u00ebto raste, studimi i superioritetit nuk \u00ebsht\u00eb zgjidhja m\u00eb e duhur. Megjithat\u00eb, shumica e specialist\u00ebve p\u00ebrdorin default k\u00ebt\u00eb metod\u00eb n\u00eb situata t\u00eb tilla. Ne kryejm\u00eb me kujdes nj\u00eb eksperiment p\u00ebr t\u00eb p\u00ebrcaktuar sakt\u00ebsisht madh\u00ebsin\u00eb e efektit. N\u00ebse do t\u00eb ishte e v\u00ebrtet\u00eb se versionet A dhe B funksionojn\u00eb gjithashtu n\u00eb nj\u00eb m\u00ebnyr\u00eb shum\u00eb t\u00eb ngjashme, ka mund\u00ebsi q\u00eb ne t\u00eb mos arrijm\u00eb t\u00eb refuzojm\u00eb hipotez\u00ebn nul. A arrijm\u00eb n\u00eb p\u00ebrfundimin se A dhe B funksionojn\u00eb n\u00eb p\u00ebrgjith\u00ebsi nj\u00ebsoj? Jo! Mungesa e mund\u00ebsis\u00eb p\u00ebr t\u00eb refuzuar hipotez\u00ebn nul dhe pranimi i saj nuk jan\u00eb e nj\u00ebjta gj\u00eb.<\/p>\n

Llogarit\u00eb e madh\u00ebsis\u00eb s\u00eb mostrav\u00ebs (t\u00eb cilat ju sigurisht i keni b\u00ebr\u00eb) zakonisht kryhen me kufij m\u00eb t\u00eb strikt p\u00ebr gabimin e par\u00eb (probabiliteti i refuzimit t\u00eb gabuar t\u00eb hipotez\u00ebs null, shpesh i quajtur alfa) se sa p\u00ebr gabimin e dyt\u00eb (probabiliteti i mosrefuzimit t\u00eb hipotez\u00ebs null, duke supozuar se hipoteza null \u00ebsht\u00eb e gabuar, shpesh i quajtur beta). Nj\u00eb vler\u00eb tipike p\u00ebr alfa \u00ebsht\u00eb 0,05, nd\u00ebrsa nj\u00eb vler\u00eb tipike p\u00ebr beta \u00ebsht\u00eb 0,20, q\u00eb korrespondon me fuqin\u00eb statistikore 0,80. Kjo do t\u00eb thot\u00eb se ne mund t\u00eb mos zbulojm\u00eb ndikimin e v\u00ebrtet\u00eb t\u00eb madh\u00ebsis\u00eb q\u00eb kemi p\u00ebrshkruar n\u00eb llogaritjet tona t\u00eb fuqis\u00eb, me nj\u00eb probabilitet prej 20% dhe kjo \u00ebsht\u00eb nj\u00eb boshll\u00ebk mjaft serioz n\u00eb informacion. Si nj\u00eb shembull, le t\u00eb shqyrtojm\u00eb k\u00ebto hipoteza:<\/p>\n

\"Kur
\n
\nH0: \u00e7anta ime nuk \u00ebsht\u00eb n\u00eb dhom\u00ebn time (3)
\nH1: \u00e7anta ime \u00ebsht\u00eb n\u00eb dhom\u00ebn time (4)<\/i><\/p>\n

N\u00ebse kam k\u00ebrkuar dhom\u00ebn time dhe kam gjetur \u00e7ant\u00ebn time \u2014 shk\u00eblqyesh\u00ebm, mund t\u00eb heq dor\u00eb nga hipoteza bosh. Por n\u00ebse kam kontrolluar dhom\u00ebn dhe nuk kam gjetur \u00e7ant\u00ebn time (figura 1), \u00e7far\u00eb p\u00ebrfundimi duhet t\u00eb arrij? A jam i sigurt se ajo nuk \u00ebsht\u00eb atje? A kam k\u00ebrkuar mjaftuesh\u00ebm? \u00c7far\u00eb n\u00ebse kam k\u00ebrkuar vet\u00ebm 80% t\u00eb dhom\u00ebs? T\u00eb nxjerr\u00ebsh p\u00ebrfundimin se \u00e7anta nuk \u00ebsht\u00eb n\u00eb dhom\u00eb do t\u00eb ishte nj\u00eb vendim i nxituar. Nuk \u00ebsht\u00eb \u00e7udi q\u00eb nuk mund ta \u201cpranojm\u00eb hipotez\u00ebn bosh\u201d.
\n\"Kur
\nZona q\u00eb kemi kontrolluar
\nNuk e gjet\u00ebm \u00e7ant\u00ebn \u2014 a duhet t\u00eb pranojm\u00eb hipotez\u00ebn bosh?<\/i><\/p>\n

Figura 1. K\u00ebrkimi i 80% t\u00eb dhom\u00ebs \u00ebsht\u00eb p\u00ebraf\u00ebrsisht si t\u00eb realizosh nj\u00eb studim me fuqin\u00eb 80%. N\u00ebse nuk e gjet\u00ebt \u00e7ant\u00ebn duke kontrolluar 80% t\u00eb dhom\u00ebs, a mund t\u00eb nxjerr\u00ebsh p\u00ebrfundimin se ajo nuk \u00ebsht\u00eb atje?<\/i> <\/p>\n

\u00c7far\u00eb duhet t\u00eb b\u00ebj\u00eb specialisti i t\u00eb dh\u00ebnave n\u00eb k\u00ebt\u00eb situat\u00eb? Mund t\u00eb rris\u00ebsh ndjesh\u00ebm fuqin\u00eb e studimit, por at\u00ebher\u00eb do t\u00eb duhet nj\u00eb most\u00ebr shum\u00eb m\u00eb e madhe, dhe rezultati do t\u00eb jet\u00eb akoma i pak\u00ebnaqsh\u00ebm.<\/p>\n

Fatkeq\u00ebsisht, k\u00ebto probleme jan\u00eb studiuar prej koh\u00ebsh n\u00eb bot\u00ebn e k\u00ebrkimeve klinike. Ila\u00e7i B \u00ebsht\u00eb m\u00eb i lir\u00eb se ila\u00e7i A; pritet q\u00eb ila\u00e7i B t\u00eb shkaktoj\u00eb m\u00eb pak efekte an\u00ebsore se ila\u00e7i A; ila\u00e7i B \u00ebsht\u00eb m\u00eb i leht\u00eb p\u00ebr tu transportuar, sepse nuk ka nevoj\u00eb t\u00eb ruhet n\u00eb frigorifer, nd\u00ebrsa ila\u00e7i A duhet. Le t\u00eb verifikojm\u00eb hipotez\u00ebn e efektivitetit t\u00eb pakt\u00eb. Kjo \u00ebsht\u00eb e nevojshme p\u00ebr t\u00eb treguar se versioni B \u00ebsht\u00eb po aq i mir\u00eb sa versioni A \u2014 t\u00eb pakt\u00ebn brenda nj\u00eb kufiri t\u00eb paracaktuar t\u00eb \u201cefektivitetit t\u00eb pakt\u00eb\u201d, \u0394. Pak m\u00eb von\u00eb do t\u00eb flasim m\u00eb n\u00eb detaje se si t\u00eb p\u00ebrcaktojm\u00eb k\u00ebt\u00eb kufi. Por tani le t\u00eb supozojm\u00eb se kjo \u00ebsht\u00eb diferenca minimale q\u00eb ka r\u00ebnd\u00ebsi praktike (n\u00eb kontekstin e provave klinike, zakonisht quhet r\u00ebnd\u00ebsi klinike).<\/p>\n

Hipotezat mbi efektivitetin e pakt\u00eb e p\u00ebrmbysin gjith\u00e7ka:<\/p>\n

\"Kur
\n
\nTaniher\u00eb, p\u00ebrpara se t\u00eb supozojm\u00eb se nuk ka ndryshim, ne supozojm\u00eb se versione B \u00ebsht\u00eb m\u00eb e keqe se versioni A, dhe do t\u00eb mbetemi n\u00eb k\u00ebt\u00eb supozim derisa t\u00eb demonstrojm\u00eb se nuk \u00ebsht\u00eb k\u00ebshtu. Kjo \u00ebsht\u00eb pik\u00ebrisht ajo koh\u00eb kur ka kuptim t\u00eb p\u00ebrdoret testimi i hipotez\u00ebs nj\u00ebansh\u00ebm! N\u00eb praktik\u00eb, kjo mund t\u00eb b\u00ebhet duke nd\u00ebrtuar nj\u00eb interval besimi dhe duke p\u00ebrcaktuar n\u00ebse intervali \u00ebsht\u00eb v\u00ebrtet m\u00eb i madh se \u0394 (Figura 2).
\n\"Kur
\n<\/p>\n

Zgjedhja e \u0394<\/h3>\n

\nSi t\u00eb zgjidhni sakt\u00ebsisht \u0394? Procesi i zgjedhjes s\u00eb \u0394 p\u00ebrfshin justifikim statistik dhe vler\u00ebsim tematik. N\u00eb bot\u00ebn e k\u00ebrkimeve klinike ekzistojn\u00eb rekomandime normative, sipas t\u00eb cilave delta duhet t\u00eb p\u00ebrfaq\u00ebsoj\u00eb diferenc\u00ebn m\u00eb t\u00eb vog\u00ebl klinikisht t\u00eb r\u00ebnd\u00ebsishme \u2014 nj\u00eb ndryshim q\u00eb do t\u00eb ket\u00eb r\u00ebnd\u00ebsi n\u00eb praktik\u00eb. Ja nj\u00eb citat nga udh\u00ebzimi evropian, p\u00ebr t\u00eb cilin mund t\u00eb kontrolloni veten: \"N\u00ebse ndryshimi \u00ebsht\u00eb zgjedhur si\u00e7 duhet, intervali i besimit, i cili \u00ebsht\u00eb plot\u00ebsisht mes \u2013\u2206 dhe 0..., ende \u00ebsht\u00eb i mjaftuesh\u00ebm p\u00ebr t\u00eb demonstruar efikasitetin e pakt\u00eb. N\u00ebse ky rezultat nuk duket i pranuesh\u00ebm, kjo do t\u00eb thot\u00eb se \u2206 nuk \u00ebsht\u00eb zgjedhur si\u00e7 duhet.\"<\/p>\n

Delta p\u00ebrfundoj nuk duhet t\u00eb tejkaloj\u00eb madh\u00ebsin\u00eb e efektit t\u00eb versionit A n\u00eb raport me kontrollin e v\u00ebrtet\u00eb (placebo \/ munges\u00eb trajtimi), pasi kjo na \u00e7on n\u00eb faktin se versioni B \u00ebsht\u00eb m\u00eb i keq se kontrolli i v\u00ebrtet\u00eb, dhe nj\u00ebher\u00ebsh tregon \"jo m\u00eb pak efikasitet\". Supozojm\u00eb se kur u prezantua versioni A, n\u00eb vendin e tij kishte versionin 0 ose funksioni nuk ekzistonte fare (shih figur\u00ebn 3). <\/p>\n

Sipas rezultateve t\u00eb hetimit t\u00eb hipotez\u00ebs mbi superioritetin, u zbulua madh\u00ebsia e efektit E (dometh\u00ebn\u00eb supozim \u03bc^A\u2212\u03bc^0=E). Tani A \u00ebsht\u00eb standardi yn\u00eb i ri, dhe ne duam t\u00eb sigurohemi se B nuk i jep pas A. Nj\u00eb m\u00ebnyr\u00eb tjet\u00ebr p\u00ebr t\u00eb shkruar \u03bcB\u2212\u03bcA\u2264\u2212\u0394 (hipoteza zero) \u00ebsht\u00eb \u03bcB\u2264\u03bcA\u2212\u0394. N\u00ebse ne pranojm\u00eb se e b\u00ebrra \u00ebsht\u00eb e barabart\u00eb ose e tejkalon E, at\u00ebher\u00eb \u03bcB \u2264 \u03bcA\u2212E \u2264 placebo. Tani ne shohim se vler\u00ebsimi yn\u00eb p\u00ebr \u03bcB plot\u00ebsisht tejkalon \u03bcA\u2212E, gj\u00eb q\u00eb n\u00eb m\u00ebnyr\u00eb t\u00eb plot\u00eb v\u00ebrteton hipotez\u00ebn zero dhe lejon t\u00eb p\u00ebrfundojm\u00eb se B nuk i jep pas A, por nj\u00ebkoh\u00ebsisht \u03bcB mund t\u00eb jet\u00eb \u2264 \u03bc placebo, dhe kjo nuk \u00ebsht\u00eb ajo q\u00eb na nevojitet. (figur\u00eb 3).<\/p>\n

\"Kur
\nFigura 3. Demonstrimi i rreziqeve t\u00eb zgjedhjes s\u00eb kufirit t\u00eb efikasitetit t\u00eb pandash\u00ebm. N\u00ebse kufiri \u00ebsht\u00eb shum\u00eb i madh, mund t\u00eb konkludojm\u00eb se B nuk i p\u00ebrshtatet A, por gjithashtu \u00ebsht\u00eb i pandash\u00ebm nga placebo. Nuk do ta z\u00ebvend\u00ebsojm\u00eb nj\u00eb ila\u00e7 q\u00eb \u00ebsht\u00eb duksh\u00ebm m\u00eb efektiv se placebo (A) me nj\u00eb ila\u00e7 q\u00eb ka t\u00eb nj\u00ebjtin efikasitet si placebo. <\/i><\/p>\n

Zgjedhja \u03b1 <\/h3>\n

\nT\u00eb kalojm\u00eb n\u00eb zgjedhjen e \u03b1. Mund t\u00eb p\u00ebrdorim vler\u00ebn standarde \u03b1 = 0,05, por kjo nuk \u00ebsht\u00eb krejt\u00ebsisht e ndershme. Sikur kur blini di\u00e7ka n\u00eb internet dhe p\u00ebrdorni disa kode t\u00eb zbritjes menj\u00ebher\u00eb, ndon\u00ebse ato nuk duhet t\u00eb akumulojn\u00eb \u2014 thjesht zhvilluesi b\u00ebri nj\u00eb gabim, dhe ju kaluat me t\u00eb. Sip\u00ebrfaqja duhet t\u00eb jet\u00eb e barabart\u00eb me gjysm\u00ebn e vler\u00ebs s\u00eb \u03b1 q\u00eb p\u00ebrdoret n\u00eb verifikimin e hipotez\u00ebs p\u00ebr superioritetin, do t\u00eb thot\u00eb 0,05 \/ 2 = 0,025.<\/p>\n

Madh\u00ebsia e mostr\u00ebs <\/h3>\n

\nSi t\u00eb vler\u00ebsoni madh\u00ebsin\u00eb e mostr\u00ebs? N\u00ebse mendoni se diferenca reale e mesatareve midis A dhe B \u00ebsht\u00eb 0, at\u00ebher\u00eb llogaritja e madh\u00ebsis\u00eb s\u00eb mostr\u00ebs do t\u00eb jet\u00eb e nj\u00ebjt\u00eb me at\u00eb q\u00eb p\u00ebrdoret n\u00eb verifikimin e hipotez\u00ebs p\u00ebr superioritetin, p\u00ebrve\u00e7 se z\u00ebvend\u00ebsoni madh\u00ebsin\u00eb e efektit me kufirin e efikasitetit t\u00eb pandash\u00ebm, me kusht q\u00eb t\u00eb p\u00ebrdorni efikasiteti i barabart\u00eb = 1\/2 t\u00eb superioritetit <\/b>(\u03b1non-inferioriteti = 1\/2 \u03b1superioriteti). N\u00ebse keni arsye t\u00eb besoni se opsioni B mund t\u00eb jet\u00eb pak m\u00eb i keq se opsioni A, por d\u00ebshironi t\u00eb provoni se ai nuk \u00ebsht\u00eb m\u00eb i keq se \u0394, at\u00ebher\u00eb ju keni fat! N\u00eb fakt, kjo zvog\u00eblon madh\u00ebsin\u00eb e mostr\u00ebs tuaj, sepse \u00ebsht\u00eb m\u00eb e leht\u00eb t\u00eb demonstroni se B \u00ebsht\u00eb m\u00eb i keq se A, n\u00ebse ju n\u00eb fakt besoni se ai \u00ebsht\u00eb pak m\u00eb i keq dhe jo ekvivalent.<\/p>\n

Shembulli me vendimin <\/h3>\n

\nSupozoni se d\u00ebshironi t\u00eb kaloni n\u00eb versionin B me kushtin q\u00eb ai t\u00eb jet\u00eb m\u00eb i keq se versioni A jo m\u00eb shum\u00eb se 0.1 pik\u00eb n\u00eb nj\u00eb shkall\u00eb t\u00eb 5 pikave t\u00eb k\u00ebnaq\u00ebsis\u00eb s\u00eb klient\u00ebve\u2026 Le t\u00eb qasemi k\u00ebt\u00eb problem duke p\u00ebrdorur hipotez\u00ebn e superioritetit.<\/p>\n

P\u00ebr t\u00eb testuar hipotez\u00ebn e superioritetit, do t\u00eb llogaritnim madh\u00ebsin\u00eb e mostr\u00ebs si m\u00eb posht\u00eb:<\/p>\n

\"Kur
\n
\nPra ndodhi q\u00eb n\u00eb grupin tuaj t\u00eb ket\u00eb 2103 v\u00ebzhgime, mund t\u00eb jeni 90% t\u00eb sigurt q\u00eb do t\u00eb zbuloni nj\u00eb efekt me madh\u00ebsi 0,10 ose m\u00eb shum\u00eb. Por n\u00ebse vlera 0,10 \u00ebsht\u00eb shum\u00eb e madhe p\u00ebr ju, ndoshta nuk duhet t\u00eb kontrolloni hipotez\u00ebn p\u00ebr superioritetin. Mund t\u00eb vendosni t\u00eb realizoni nj\u00eb studim p\u00ebr nj\u00eb efekt m\u00eb t\u00eb vog\u00ebl, si 0,05. N\u00eb k\u00ebt\u00eb rast, do t'ju duhen 8407 v\u00ebzhgime, q\u00eb do t\u00eb thot\u00eb se mostra do t\u00eb rritet gati kat\u00ebrfish. Por \u00e7far\u00eb n\u00ebse mbetemi me madh\u00ebsin\u00eb ton\u00eb fillestare t\u00eb mostr\u00ebs, por rrisim fuqin\u00eb n\u00eb 0,99, n\u00eb m\u00ebnyr\u00eb q\u00eb t\u00eb mos kemi dyshime n\u00ebse marrim nj\u00eb rezultat pozitiv? N\u00eb k\u00ebt\u00eb rast, n p\u00ebr nj\u00eb grup do t\u00eb jet\u00eb 3676, q\u00eb \u00ebsht\u00eb m\u00eb mir\u00eb, por rrit edhe m\u00eb shum\u00eb madh\u00ebsin\u00eb e mostr\u00ebs mbi 50%. Dhe p\u00ebrfundimisht, ne p\u00ebrs\u00ebri nuk do t\u00eb mund t\u00eb hedhim posht\u00eb hipotez\u00ebn e zeros, por as nuk do t\u00eb marrim nj\u00eb p\u00ebrgjigje p\u00ebr pyetjen ton\u00eb.<\/p>\n

\u00c7far\u00eb n\u00ebse p\u00ebr k\u00ebt\u00eb n\u00eb vend t\u00eb k\u00ebsaj kontrollojm\u00eb hipotez\u00ebn p\u00ebr nj\u00eb efikasitet jo m\u00eb t\u00eb vog\u00ebl? <\/p>\n

\"Kur
\n
\nMadh\u00ebsia e mostr\u00ebs do t\u00eb llogaritet sipas t\u00eb nj\u00ebjt\u00ebs formul\u00eb p\u00ebrve\u00e7 denominatorit.
\nDallimet me formul\u00ebn e p\u00ebrdorur p\u00ebr kontrollimin e hipotez\u00ebs p\u00ebr superioritetin jan\u00eb si n\u00eb vazhdim:<\/p>\n

\u2014 Z1\u2212\u03b1\/2 z\u00ebvend\u00ebsohet me Z1\u2212\u03b1, por n\u00ebse e b\u00ebni gjith\u00e7ka sipas rregullave, z\u00ebvend\u00ebsoni \u03b1 = 0,05 me \u03b1 = 0,025, pra, \u00ebsht\u00eb i nj\u00ebjti num\u00ebr (1,96)<\/p>\n

\u2014 n\u00eb em\u00ebrues shfaqet (\u03bcB\u2212\u03bcA)<\/p>\n

\u2014 \u03b8 (magnitude e efektit) z\u00ebvend\u00ebsohet me \u0394 (kufi i efektshm\u00ebris\u00eb s\u00eb pakt\u00eb)<\/p>\n

N\u00ebse supozojm\u00eb se \u00b5B = \u00b5A, at\u00ebher\u00eb (\u00b5B \u2212 \u00b5A) = 0 dhe llogaritja e madh\u00ebsis\u00eb s\u00eb mostr\u00ebs p\u00ebr kufirin e efektshm\u00ebris\u00eb s\u00eb pakt\u00eb \u00ebsht\u00eb pik\u00ebrisht ajo q\u00eb do t\u00eb kishim marr\u00eb duke llogaritur superioritetin p\u00ebr magnitud\u00ebn e efektit 0,1, shk\u00eblqyesh\u00ebm! Mund t\u00eb realizojm\u00eb nj\u00eb studim me t\u00eb nj\u00ebjtin p\u00ebrmas\u00eb me hipoteza t\u00eb ndryshme dhe me nj\u00eb qasje tjet\u00ebr p\u00ebr arsyetimin, dhe do t\u00eb marrim p\u00ebrgjigjen p\u00ebr pyetjen q\u00eb na intereson me t\u00eb v\u00ebrtet\u00eb.<\/p>\n

Tani le t\u00eb supozojm\u00eb se n\u00eb t\u00eb v\u00ebrtet\u00eb nuk mendojm\u00eb se \u00b5B = \u00b5A dhe
\nmendojm\u00eb se \u00b5B \u00ebsht\u00eb pak m\u00eb i dob\u00ebt, ndoshta me 0,01 nj\u00ebsi. Kjo e rrit em\u00ebruesin ton\u00eb, duke zvog\u00ebluar madh\u00ebsin\u00eb e mostr\u00ebs p\u00ebr grupin n\u00eb 1737.<\/p>\n

\u00c7far\u00eb do ndodh\u00eb n\u00ebse versioni B \u00ebsht\u00eb n\u00eb t\u00eb v\u00ebrtet\u00eb m\u00eb i mir\u00eb se versioni A? Ne hedhim posht\u00eb hipotez\u00ebn zero q\u00eb B \u00ebsht\u00eb m\u00eb keq se A p\u00ebr m\u00eb shum\u00eb se \u0394 dhe pranojm\u00eb hipotez\u00ebn alternative q\u00eb B, n\u00ebse \u00ebsht\u00eb m\u00eb keq, nuk \u00ebsht\u00eb m\u00eb keq se \u0394 dhe mund t\u00eb jet\u00eb m\u00eb i mir\u00eb. Provoni ta p\u00ebrfshini k\u00ebt\u00eb p\u00ebrfundim n\u00eb nj\u00eb prezantim nd\u00ebrfunksional dhe shihni \u00e7far\u00eb ndodh (seriozisht, provoni). N\u00eb nj\u00eb situat\u00eb kur duhet t\u00eb orientoheni n\u00eb perspektiv\u00eb, askush nuk d\u00ebshiron t\u00eb pajtohet me \"m\u00eb keq jo m\u00eb shum\u00eb se \u0394 dhe ndoshta m\u00eb mir\u00eb\".<\/p>\n

N\u00eb k\u00ebt\u00eb rast, ne mund t\u00eb kryejm\u00eb nj\u00eb studim q\u00eb quhet shum\u00eb shkurt \"verifikimi i hipotez\u00ebs se nj\u00eb nga mund\u00ebsit\u00eb tejkalon tjetr\u00ebn ose i n\u00ebnshtrohet asaj\". Ajo p\u00ebrdor dy grupe hipotezash:<\/p>\n

Grupi i par\u00eb (i nj\u00ebjt\u00eb si n\u00eb verifikimin e hipotez\u00ebs p\u00ebr efikasitetin e barabart\u00eb): <\/p>\n

\"Kur
\n
\nGrupi i dyt\u00eb (i nj\u00ebjt\u00eb si n\u00eb verifikimin e hipotez\u00ebs p\u00ebr mbizot\u00ebrim):<\/p>\n

\"Kur
\n
\nNe e nevojshme t\u00eb testojm\u00eb hipotez\u00ebn e dyt\u00eb vet\u00ebm n\u00ebse hipoteza e par\u00eb \u00ebsht\u00eb refuzuar. Gjat\u00eb testimit t\u00eb renditsh\u00ebm, ne ruajm\u00eb nivelin e p\u00ebrgjithsh\u00ebm t\u00eb gabimeve t\u00eb natyr\u00ebs s\u00eb par\u00eb (\u03b1). N\u00eb praktik\u00eb, kjo mund t\u00eb arrihet duke krijuar nj\u00eb interval besimi prej 95% p\u00ebr diferenc\u00ebn midis mesatareve dhe kontrolluar n\u00ebse i gjith intervali kalon -\u0394. N\u00ebse intervali nuk e kalon -\u0394, ne nuk mund t\u00eb refuzojm\u00eb vler\u00ebn zero dhe do t\u00eb ndalemi. N\u00ebse i gjith\u00eb intervali n\u00eb t\u00eb v\u00ebrtet\u00eb kalon -\u0394, ne do t\u00eb vazhdojm\u00eb dhe do t\u00eb shohim n\u00ebse intervali p\u00ebrmban 0. <\/p>\n

Ekziston nj\u00eb lloj tjet\u00ebr k\u00ebrkimesh, t\u00eb cilin nuk e kemi diskutuar \u2013 k\u00ebrkime ekvivalent\u00ebsie. <\/p>\n

K\u00ebto lloje studimesh mund t\u00eb z\u00ebvend\u00ebsohen nga studime p\u00ebr t\u00eb verifikuar hipotez\u00ebn e mosferkimit t\u00eb efikasitetit dhe anasjelltas, por ato kan\u00eb nj\u00eb dallim t\u00eb r\u00ebnd\u00ebsish\u00ebm. Provimi p\u00ebr t\u00eb verifikuar hipotez\u00ebn e mosferkimit t\u00eb efikasitetit ka p\u00ebr q\u00ebllim t\u00eb tregoj\u00eb se varianti B \u00ebsht\u00eb s\u00eb paku po aq i mir\u00eb sa A. Nd\u00ebrsa studimi i ekvivalenc\u00ebs ka p\u00ebr q\u00ebllim t\u00eb tregoj\u00eb se varianti B \u00ebsht\u00eb s\u00eb paku po aq i mir\u00eb sa A, dhe varianti A po ashtu i mir\u00eb si B, q\u00eb \u00ebsht\u00eb m\u00eb e komplikuar. N\u00eb thelb, ne po p\u00ebrpiqemi t\u00eb p\u00ebrcaktojm\u00eb n\u00ebse gjith\u00eb intervali i besimit p\u00ebr diferenc\u00ebn e mesatareve \u00ebsht\u00eb midis \u2212\u0394 dhe \u0394. K\u00ebto studime k\u00ebrkojn\u00eb nj\u00eb madh\u00ebsi m\u00eb t\u00eb madhe t\u00eb mostr\u00ebs dhe kryhen m\u00eb rrall\u00eb. Prandaj, her\u00ebn tjet\u00ebr kur t\u00eb realizoni nj\u00eb studim, n\u00ebse detyra juaj kryesore \u00ebsht\u00eb t\u00eb siguroheni q\u00eb versioni i ri nuk \u00ebsht\u00eb m\u00eb i dob\u00ebt, mos pranoni \"pamund\u00ebsin\u00eb e p\u00ebrg\u00ebnjeshtrimit t\u00eb hipotez\u00ebs s\u00eb null\u00ebs\". N\u00ebse d\u00ebshironi t\u00eb provoni nj\u00eb hipotez\u00eb v\u00ebrtet t\u00eb r\u00ebnd\u00ebsishme, shqyrtoni opsione t\u00eb ndryshme.
\n
Burimi: habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u0421\u0442\u0430\u0442\u044c\u044f \u043e\u0442 \u043a\u043e\u043c\u0430\u043d\u0434\u044b Stitch Fix \u043f\u0440\u0435\u0434\u043b\u0430\u0433\u0430\u0435\u0442 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u044c \u043f\u043e\u0434\u0445\u043e\u0434 \u043a\u043b\u0438\u043d\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u043d\u0438\u0439 \u043d\u0435 \u043c\u0435\u043d\u044c\u0448\u0435\u0439 \u044d\u0444\u0444\u0435\u043a\u0442\u0438\u0432\u043d\u043e\u0441\u0442\u0438 (non-inferiority trials) \u0432 \u043c\u0430\u0440\u043a\u0435\u0442\u0438\u043d\u0433\u043e\u0432\u044b\u0445 \u0438 \u043f\u0440\u043e\u0434\u0443\u043a\u0442\u043e\u0432\u044b\u0445 A\/B \u0442\u0435\u0441\u0442\u0430\u0445. \u0422\u0430\u043a\u043e\u0439 \u043f\u043e\u0434\u0445\u043e\u0434 \u0434\u0435\u0439\u0441\u0442\u0432\u0438\u0442\u0435\u043b\u044c\u043d\u043e \u043f\u0440\u0438\u043c\u0435\u043d\u0438\u043c, \u043a\u043e\u0433\u0434\u0430 \u043c\u044b \u0442\u0435\u0441\u0442\u0438\u0440\u0443\u0435\u043c \u043d\u043e\u0432\u043e\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435, \u0438\u043c\u0435\u044e\u0449\u0435\u0435 \u043f\u0440\u0435\u0438\u043c\u0443\u0449\u0435\u0441\u0442\u0432\u0430, \u043d\u0435\u0438\u0437\u043c\u0435\u0440\u044f\u0435\u043c\u044b\u0435 \u0442\u0435\u0441\u0442\u0430\u043c\u0438. \u0421\u0430\u043c\u044b\u0439 \u043f\u0440\u043e\u0441\u0442\u043e\u0439 \u043f\u0440\u0438\u043c\u0435\u0440 \u2014 \u0441\u043d\u0438\u0436\u0435\u043d\u0438\u0435 \u043a\u043e\u0441\u0442\u043e\u0432. \u041d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u0430\u0432\u0442\u043e\u043c\u0430\u0442\u0438\u0437\u0438\u0440\u0443\u0435\u043c \u043f\u0440\u043e\u0446\u0435\u0441\u0441 \u043d\u0430\u0437\u043d\u0430\u0447\u0435\u043d\u0438\u044f \u043f\u0435\u0440\u0432\u043e\u0433\u043e \u0443\u0440\u043e\u043a\u0430, \u043d\u043e \u043d\u0435 \u0445\u043e\u0442\u0438\u043c \u0441\u0438\u043b\u044c\u043d\u043e \u0443\u0440\u043e\u043d\u0438\u0442\u044c \u0441\u043a\u0432\u043e\u0437\u043d\u0443\u044e \u043a\u043e\u043d\u0432\u0435\u0440\u0441\u0438\u044e. \u0418\u043b\u0438 \u0442\u0435\u0441\u0442\u0438\u0440\u0443\u0435\u043c […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":26311,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-34958","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-novosti-interneta"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Kur duhet t\u00eb verifikohet hipoteza p\u00ebr barazpesh\u00eb efektiviteti? | ProHoster","description":"Artikulli nga ekipi i Stitch Fix sugjeron p\u00ebrdorimin e nj\u00eb qasjeje t\u00eb k\u00ebrkimeve klinike t\u00eb barazpesh\u00ebs (non-inferiority trials) n\u00eb testet A\/B p\u00ebr marketingun dhe produktet. Kjo qasje \u00ebsht\u00eb v\u00ebrtet e aplikueshme kur testojm\u00eb nj\u00eb zgjidhje t\u00eb re, e cila ka p\u00ebrfitime q\u00eb nuk maten nga testet. Nj\u00eb shembull i thjesht\u00eb \u00ebsht\u00eb ulja e kostove. P\u00ebr shembull, automatizojm\u00eb procesin e caktimit t\u00eb m\u00ebsimit t\u00eb par\u00eb, por nuk duam t\u00eb d\u00ebmtojm\u00eb ndjesh\u00ebm konvertimin p\u00ebrfundimtar. Ose po testojm\u00eb","canonical_url":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/kogda-stoit-proveryat-gipotezu-o-ne-menshej-effektivnosti","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"sq_AL","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u041a\u043e\u0433\u0434\u0430 \u0441\u0442\u043e\u0438\u0442 \u043f\u0440\u043e\u0432\u0435\u0440\u044f\u0442\u044c \u0433\u0438\u043f\u043e\u0442\u0435\u0437\u0443 \u043e \u043d\u0435 \u043c\u0435\u043d\u044c\u0448\u0435\u0439 \u044d\u0444\u0444\u0435\u043a\u0442\u0438\u0432\u043d\u043e\u0441\u0442\u0438? | ProHoster","og:description":"\u0421\u0442\u0430\u0442\u044c\u044f \u043e\u0442 \u043a\u043e\u043c\u0430\u043d\u0434\u044b Stitch Fix \u043f\u0440\u0435\u0434\u043b\u0430\u0433\u0430\u0435\u0442 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u044c \u043f\u043e\u0434\u0445\u043e\u0434 \u043a\u043b\u0438\u043d\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0438\u0441\u0441\u043b\u0435\u0434\u043e\u0432\u0430\u043d\u0438\u0439 \u043d\u0435 \u043c\u0435\u043d\u044c\u0448\u0435\u0439 \u044d\u0444\u0444\u0435\u043a\u0442\u0438\u0432\u043d\u043e\u0441\u0442\u0438 (non-inferiority trials) \u0432 \u043c\u0430\u0440\u043a\u0435\u0442\u0438\u043d\u0433\u043e\u0432\u044b\u0445 \u0438 \u043f\u0440\u043e\u0434\u0443\u043a\u0442\u043e\u0432\u044b\u0445 A\/B \u0442\u0435\u0441\u0442\u0430\u0445. \u0422\u0430\u043a\u043e\u0439 \u043f\u043e\u0434\u0445\u043e\u0434 \u0434\u0435\u0439\u0441\u0442\u0432\u0438\u0442\u0435\u043b\u044c\u043d\u043e \u043f\u0440\u0438\u043c\u0435\u043d\u0438\u043c, \u043a\u043e\u0433\u0434\u0430 \u043c\u044b \u0442\u0435\u0441\u0442\u0438\u0440\u0443\u0435\u043c \u043d\u043e\u0432\u043e\u0435 \u0440\u0435\u0448\u0435\u043d\u0438\u0435, \u0438\u043c\u0435\u044e\u0449\u0435\u0435 \u043f\u0440\u0435\u0438\u043c\u0443\u0449\u0435\u0441\u0442\u0432\u0430, \u043d\u0435\u0438\u0437\u043c\u0435\u0440\u044f\u0435\u043c\u044b\u0435 \u0442\u0435\u0441\u0442\u0430\u043c\u0438. \u0421\u0430\u043c\u044b\u0439 \u043f\u0440\u043e\u0441\u0442\u043e\u0439 \u043f\u0440\u0438\u043c\u0435\u0440 \u2014 \u0441\u043d\u0438\u0436\u0435\u043d\u0438\u0435 \u043a\u043e\u0441\u0442\u043e\u0432. \u041d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u0430\u0432\u0442\u043e\u043c\u0430\u0442\u0438\u0437\u0438\u0440\u0443\u0435\u043c \u043f\u0440\u043e\u0446\u0435\u0441\u0441 \u043d\u0430\u0437\u043d\u0430\u0447\u0435\u043d\u0438\u044f \u043f\u0435\u0440\u0432\u043e\u0433\u043e \u0443\u0440\u043e\u043a\u0430, \u043d\u043e \u043d\u0435 \u0445\u043e\u0442\u0438\u043c \u0441\u0438\u043b\u044c\u043d\u043e \u0443\u0440\u043e\u043d\u0438\u0442\u044c \u0441\u043a\u0432\u043e\u0437\u043d\u0443\u044e \u043a\u043e\u043d\u0432\u0435\u0440\u0441\u0438\u044e. \u0418\u043b\u0438 \u0442\u0435\u0441\u0442\u0438\u0440\u0443\u0435\u043c","og:url":"https:\/\/prohoster.info\/sq\/blog\/novosti-interneta\/kogda-stoit-proveryat-gipotezu-o-ne-menshej-effektivnosti","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2019-10-31T19:01:31+00:00","article:modified_time":"2019-10-31T19:01:31+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"34958","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-21 21:15:19","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-03-01 02:12:27","updated":"2026-01-21 21:15:19","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/34958","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/comments?post=34958"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/posts\/34958\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/media\/26311"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/media?parent=34958"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/categories?post=34958"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/sq\/wp-json\/wp\/v2\/tags?post=34958"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}