Variability in the analysis of a single neuroimaging dataset by many teams
Botvinik-Nezer R., Holzmeister F., Camerer C., Dreber A., Huber J., Johannesson M., Kirchler M., Iwanir R., Mumford J., Adcock A., Avesani P., Baczkowski B., Bajracharya A., Bakst L., Ball S., Barilari M., Bault N., Beaton D., Beitner J., Benoit R., Berkers R., Bhanji J., Biswal B., Bobadilla-Suarez S., Bortolini T., Bottenhorn K., Bowring A., Braem S., Brooks H., Brudner E., Calderon C., Camilleri J., Castrellon J., Cecchetti L., Cieslik E., Cole Z., Collignon O., Cox R., Cunningham W., Czoschke S., Dadi K., Davis C., De Luca A., Delgado M., Demetriou L., Dennison J., Di X., Dickie E., Dobryakova E., Donnat C., Dukart J., Duncan N., Durnez J., Eed A., Eickhoff S., Erhart A., Fontanesi L., Fricke M., Galvan A., Gau R., Genon S., Glatard T., Glerean E., Goeman J., Golowin S., González-García C., Gorgolewski K., Grady C., Green M., Guassi Moreira J., Guest O., Hakimi S., Hamilton P., Hancock R., Handjaras G., Harry B., Hawco C., Herholz P., Herman G., Heunis S., Hoffstaedter F., Hogeveen J., Holmes S., Hu C-P., Huettel S., Hughes M., Iacovella V., Iordan A., Isager P., Isik AI., Jahn A., Johnson M., Johnstone T., Joseph M., Juliano A., Kable J., Kassinopoulos M., Koba C., Kong X-Z., Koscik T., Kucukboyaci NE., Kuhl B., Kupek S., Laird A., Lamm C., Langner R., Lauharatanahirun N., Lee H., Lee S., Leemans A., Leo A., Lesage E., Li F., Li M., Lim PC., Lintz E., Liphardt S., Losecaat Vermeer A., Love B., Mack M., Malpica N., Marins T., Maumet C., McDonald K., McGuire J., Melero H., Méndez Leal A., Meyer B., Meyer K., Mihai P., Mitsis G., Moll J., Nielson D., Nilsonne G., Notter M., Olivetti E., Onicas A., Papale P., Patil K., Peelle J., Pérez A., Pischedda D., Poline J-B., Prystauka Y., Ray S., Reuter-Lorenz P., Reynolds R., Ricciardi E., Rieck J., Rodriguez-Thompson A., Romyn A., Salo T., Samanez-Larkin G., Sanz-Morales E., Schlichting M., Schultz D., Shen Q., Sheridan M., Shiguang F., Silvers J., Skagerlund K., Smith A., Smith D., Sokol-Hessner P., Steinkamp S., Tashjian S., Thirion B., Thorp J., Tinghög G., Tisdall L., Tompson S., Toro-Serey C., Torre J., Tozzi L., Truong V., Turella L., van’t Veer A., Verguts T., Vettel J., Vijayarajah S., Vo K., Wall M., Weeda W., Weis S., White D., Wisniewski D., Xifra-Porxas A., Yearling E., Yoon S., Yuan R., Yuen K., Zhang L., Zhang X., Zosky J., Nichols T., Poldrack R., Schonberg T.
Summary Data analysis workflows in many scientific domains have become increasingly complex and flexible. To assess the impact of this flexibility on functional magnetic resonance imaging (fMRI) results, the same dataset was independently analyzed by 70 teams, testing nine ex-ante hypotheses. The flexibility of analytic approaches is exemplified by the fact that no two teams chose identical workflows to analyze the data. This flexibility resulted in sizeable variation in hypothesis test results, even for teams whose statistical maps were highly correlated at intermediate stages of their analysis pipeline. Variation in reported results was related to several aspects of analysis methodology. Importantly, meta-analytic approaches that aggregated information across teams yielded significant consensus in activated regions across teams. Furthermore, prediction markets of researchers in the field revealed an overestimation of the likelihood of significant findings, even by researchers with direct knowledge of the dataset. Our findings show that analytic flexibility can have substantial effects on scientific conclusions, and demonstrate factors related to variability in fMRI. The results emphasize the importance of validating and sharing complex analysis workflows, and demonstrate the need for multiple analyses of the same data. Potential approaches to mitigate issues related to analytical variability are discussed.