1.4.7

06-effectsize.qmd

@@ -357,7 +357,7 @@ We see the effect size of the cross-over interaction (*f* = 0.25) is twice as la
 
 Just stating that you expect a 'medium' Cohen's *f* effect size for an interaction effect in your power analysis is not the best approach. Instead, start by thinking about the pattern of means and standard deviations (and for within factors, the correlation between dependent variables) and then compute the effect size from the data pattern. If you prefer not to do so by hand, you can use [Superpower](https://aaroncaldwell.us/SuperpowerBook/) [@lakens_simulationbased_2021]. This also holds for more complex designs, such as multilevel models. In these cases, it is often the case that power analyses are easier to perform with simulation-based approaches, than based on plugging a single effect size in to power analysis software [@debruine_understanding_2021].
 
-## Why Effect Sizes Selected for Significance are Inflated
+## Why Effect Sizes Selected for Significance are Inflated {#sec-effectinflated}
 
 Another way to think about this is through the concept of a **truncated distribution**. If effect sizes are only reported if the *p*-value is statistically significant, then we only have access to effect sizes that are larger than some minimal value [@taylor_bias_1996; @anderson_samplesize_2017]. In @fig-inflated only effects larger than d = 0.4 can be significant, so all effect sizes below this threshold are censored, and only effect sizes in the gray part of the distribution will be available to researchers. Without the lower part of the effect size distribution effect sizes will on average be inflated. 
 

_quarto.yml

@@ -50,6 +50,7 @@ book:
   - 14-computationalreproducibility.qmd
   - 15-researchintegrity.qmd
   - 16-confirmationbias.qmd
+  - 17-replication.qmd
   - references.qmd
   - changelog.qmd
 

changelog.qmd

@@ -4,7 +4,7 @@
 Sys.setlocale("LC_TIME", "C")
 ```
 
-The current version of this textbook is 1.4.6. 
+The current version of this textbook is 1.4.7. 
 
 This version has been compiled on `r format(Sys.Date(), '%B %d, %Y')`.
 
@@ -14,6 +14,10 @@ This page documents the changes to the textbook that were more substantial than
 
 ##### Updates {#sec-changelog}
 
+*June 20, 2024:*
+
+Added CH17
+
 *March 22, 2024:*
 
 Updated CI figure in CH7, incorporated feedback LeFoll, updated references

chapters/planned_additions

@@ -14,3 +14,11 @@ t h e selection of data i n theory-testing experiments calls for
 i n t e l l i g e n t skepticism. O n e must ask whether the selection was
 made o n the basis of legitimate practices or, consciously or otherwise,
 o n the basis of consistency w i t h a hypothesis.
+
+Th e ideal solution
+is t o encourage t h e p u b l i c a t i o n of r e l a t i v e l y long, integrated scries of
+researches. B y the time an experimenter reaches t h e end of such a
+program, he w i l l have had sufficient experience w i t h his subject
+matter to permit judgments concerning the acceptance, and reject i o n of variable data. U n f o r t u n a t e l y , the modern stress u p o n frequency
+of p u b l i c a t i o n , w i t h its economic rewards, mitigates strongly
+against such a course of action.