Short answers, pulled from the story.
Gene V. Glass coined the term "meta-analysis" in 1976 in an article published in Educational Researcher, defining it as "the analysis of analyses." Karl Pearson had used a similar aggregating approach as early as 1904, but Glass is credited with establishing the modern framework.
The first model meta-analysis was published in 1978 by Mary Lee Smith and Gene V. Glass. It examined the effectiveness of psychotherapy. Hans Eysenck responded in the same year by calling it an "exercise in mega-silliness."
Publication bias occurs when studies showing negative or insignificant results are less likely to be published, causing the available literature to skew toward positive findings. Estimates suggest up to 25% of meta-analyses in the psychological sciences may have been affected by this problem.
The fixed effect model assumes all included studies investigate the same population and weights each by the inverse of its variance, giving larger studies more influence. The random effects model accounts for variability between studies by adjusting weights based on heterogeneity, redistributing weight toward smaller studies as heterogeneity increases.
A 2011 study reviewed 29 meta-analyses covering 509 randomized controlled trials and found that 69% of trials with disclosed funding sources received industry funding, and 69% of trials with disclosed author conflicts named at least one author with financial ties to industry. Only two of the 29 meta-analyses mentioned the trial funding sources, and none disclosed author-industry financial ties.
A funnel plot is a scatter plot that places effect size on one axis and standard error on the other. When no publication bias is present the plot is symmetric, with smaller studies scattered widely at the base and larger studies clustering at the tip. An asymmetric base suggests that smaller studies pointing in one direction were disproportionately published, signaling potential publication bias.