Searchable library of reference standard deviations for calculating between-group standardized mean differences (SMDs) in randomized trials. Filter by indication, browse instruments as tiles, then choose study selection and pooling model.
Loading reference data…
The SD Reference Library provides pooled standard deviation (SD) estimates for outcome instruments commonly used in psychotherapy randomized trials. These reference SDs serve as standardizers for computing between-group standardized mean differences (SMDs) when the original trial reports only mean differences, or when a common standardizer is needed across trials.
SDs were extracted from post-treatment measurements across trials included in the Metapsy databases↗. Further details on how the catalogue was generated can be found in the published protocol↗. Pooling was performed using several model specifications:
Click an instrument tile to open the full-screen detail view. Choose Overall or Study selection, then (if needed) a dimension and level, then a pooling model. The pooled SD, confidence interval, and heterogeneity statistics appear below.
For instruments with sufficient evidence, subgroup estimates are provided across format (individual/group), recruitment (clinic/community), age group, comorbidity status, risk of bias, control condition type, and clinician-vs-self rating.
Each release is permanently archived. Use the version pill in the page header to switch versions, or share the URL after applying filters. The query string captures the active version and the current filter state, so permalinks are fully reproducible.
Reference SDs should be treated as informed priors, not as ground truth. Pooled SDs are most useful when the trial reports only mean differences, when imputing SDs for missing arms, or when computing common-metric effect sizes across heterogeneous measurement scales. Estimates should be treated with particular caution when only few estimates were available; and/or when the overall heterogeneity across contexts was high.
metapsyTools
If outcome measurements are correctly coded in the instrument variable,
you can use reference standardizers directly in
calculateEffectSizes.
Reference SDs are looked up by their individual catalogue shorthand (not the long instrument name).
Values in your Metapsy dataset’s instrument column must match this shorthand (case-insensitive).
—
Use the sd.reference argument to specify which approach should be used:
"fill" — uses the reference SDs only where values sd_arm1 / sd_arm2 are not available (NA)."override" — replace arm SDs on every matching row with the reference.The function returns both the instruments for which a matching reference standardizer was found, and the ones without a matching instrument (code) in the catalogue.
library(metapsyTools)
# instrument column must use catalogue shorthand, e.g. "BDI-II"
depressionPsyCtr %>%
checkDataFormat() %>%
checkConflicts() %>%
calculateEffectSizes(sd.reference = "fill")This is the pooled post-treatment standard deviation (SD) for the selected instrument, study selection, and pooling model. Reference SDs standardize between-group mean differences (SMDs) when trials report means without a usable trial-specific SD, or when you need a common metric across studies.
The banner shows the pooled SD with a 95% confidence interval (when estimable). Click the SD value to copy it. Subgroup and model pills summarize the active selection; use the link icon to copy a permalink to this exact estimate.
Below the banner you will find the number of contributing studies and estimates, heterogeneity (τ²), and the 95% prediction interval. Source studies are listed when references are available.
Treat reference SDs as informed priors, not ground truth; especially when only few estimates are available and/or prediction intervals are very wide.
Overall pools all studies for the instrument that pass the catalogue filters (e.g. minimum k, indication). Use this when you want a single reference SD across the full evidence base.
Study selection narrows the pool to a subgroup. Choose a dimension (such as format, recruitment setting, age group, comorbidity, risk of bias, control type, or rater) and then a level within that dimension. Only studies matching that level contribute to the estimate.
Each option shows how many estimates are available for that scope. Changing study selection updates the pooled SD, confidence interval, and heterogeneity statistics in the estimate panel.
SDs were meta-analyzed on the log scale and back-transformed. Several model specifications are available; pick the one that best matches your analysis plan.
The active model is shown as a pill on the estimate card. Switching models changes the pooled SD, interval, and heterogeneity summaries.
Each release of the SD Reference Library is permanently archived.
Permalinks are stable: ?v=N in the URL pins the version.