The Her­itage Foun­da­tion blows it, and the New York Times runs a bad head­line. Her­itage rean­a­lyzed data which shows that peo­ple who take vir­gin­i­ty pledges (a com­mon com­po­nent of absti­nence-only class­es) are 4 times more like­ly to have anal sex, 6 times more like­ly to have oral sex, have sex with­out con­tra­cep­tion, have STDs, get preg­nant, and have more abortions.

Her­itage claims they got the oppo­site results, but only by cook­ing the num­bers, and using flawed data.

The new find­ings were based on the same nation­al sur­vey used by ear­li­er stud­ies and con­duct­ed by the Depart­ment of Health and Human Ser­vices. But the authors of the new study used dif­fer­ent meth­ods of sta­tis­ti­cal analy­sis from those in an ear­li­er one that was wide­ly pub­li­cized, mak­ing direct com­par­isons difficult.

Inde­pen­dent experts called the new find­ings provoca­tive, but crit­i­cized the Her­itage team’s analy­sis as flawed and lack­ing the sta­tis­ti­cal evi­dence to back its con­clu­sions. The new find­ings have not been sub­mit­ted to a jour­nal for pub­li­ca­tion, an author said. The inde­pen­dent experts who reviewed the study said the find­ings were unlike­ly to be pub­lished in their present form.

…

The team needs to do “a lot of work” on its paper, said David Landry, a senior research asso­ciate at the Alan Guttmach­er Insti­tute in New York. He said in an inter­view that it was “a glar­ing error” to use the result of a sta­tis­ti­cal test at a 0.10 lev­el of sig­nif­i­cance when jour­nals gen­er­al­ly use a low­er and more rig­or­ous lev­el of 0.05.

Dr. John­son, a co-author, defend­ed the team’s meth­ods and said many jour­nal arti­cles used the high­er lev­el and let read­ers decide the mer­its of the findings.

Mr. Landry also crit­i­cized the Her­itage team’s reliance on self-reports of sex­u­al­ly trans­mit­ted dis­eases among those who took the pledge, say­ing that group would be less like­ly to report them. “The under­re­port­ing prob­lem is so severe that it makes that data high­ly ques­tion­able,” Mr. Landry said.

Dr. Bear­man said: “Our analy­ses showed that pledgers are less like­ly to get test­ed for S.T.D.‘s, be diag­nosed as hav­ing an S.T.D. and to see a doc­tor because they are wor­ried about hav­ing an S.T.D. Most S.T.D. infec­tions are asymp­to­matic, and there­fore, peo­ple don’t know that they have an S.T.D. unless they get test­ed. The use of self-report data for S.T.D.‘s is there­fore extreme­ly problematic.”

Mr. Landry and Dr. Freya Sonen­stein, who directs the cen­ter for ado­les­cent health at the Johns Hop­kins Bloomberg School of Pub­lic Health, urged the Her­itage team to try to pub­lish its findings.

Jour­nals may accept papers with some results at 0.1, if they have oth­er com­pelling results. Offer­ing p‑values and let­ting the read­er decide is often a wise choice. How­ev­er, if you do ten analy­ses where no sta­tis­ti­cal rela­tion­ship exists, you have a good shot at get­ting one test show­ing an effect at 0.1. I bet they ran more than 10 tests, so you have to won­der about the valid­i­ty of the tests.

And here’s the thing. If you can show that vir­gin­i­ty pledgers with an STD are less like­ly to find out that they are infect­ed, isn’t that a pret­ty good argu­ment against the pledges? After all, there is either no dif­fer­ence in infec­tion rates, or the rates are high­er in pledgers. 

Untreat­ed chlamy­dia leads to infer­til­i­ty, high­er HIV infec­tion rates, and ectopic preg­nan­cies. Ectopic preg­nan­cy can be deadly.

There­fore, vir­gin­i­ty pledges can be deadly.