diff --git a/courses/SM/2023-2024-S2/alternatives_nhst/Sad-P.webp b/courses/SM/2023-2024-S2/alternatives_nhst/Sad-P.webp new file mode 100644 index 00000000..e57dac6f Binary files /dev/null and b/courses/SM/2023-2024-S2/alternatives_nhst/Sad-P.webp differ diff --git a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.qmd b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.qmd index 0c331251..8121db8d 100644 --- a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.qmd +++ b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.qmd @@ -9,6 +9,9 @@ format: output-ext: slide.html --- +```{r child="../../../../topics/nhst_reflection/nhst_reflection.qmd", eval=TRUE} +``` + ```{r child="../../../../topics/confidence_interval/confidence_interval.qmd", eval=TRUE} ``` diff --git a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.slide.html b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.slide.html index 7cc20cb5..81a14216 100644 --- a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.slide.html +++ b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst.slide.html @@ -362,6 +362,34 @@

Alternatives to NHST

2024-02-26

+
+

The problem with P-values

+ +
+
+

There is no problem

+

The problem with P-values is that they are often misunderstood and misinterpreted. The P-value is the probability of observing a test statistic as or more extreme as the one obtained, given that the null hypothesis is true. It is not the probability that the null hypothesis is true. The P-value is not a measure of the strength of the evidence against the null hypothesis.

+
+

The mis interpretation is the problem, and not adhering to the Nayman-Pearson framework

+
+
+
+

The dance of the P-value

+ +
+
+

H0 and HA distribution

+
+ +
+
+

Confidence Interval

The confidence interval is a range of values that is likely to contain the true value of an unknown population parameter. The confidence interval is calculated from a given set of sample data. The confidence interval is used to express the uncertainty associated with a sample estimate of a population parameter.

@@ -378,15 +406,15 @@

Standard Error

  • Upperbound = \(\bar{x} + 1.96 \times SE\)
    • -
    -

    Standard Error

    +
    +

    Plot CI

    -
    +

    5 out of 100 samples

    -

    +

    @@ -429,7 +457,7 @@

    Researcher don’t know

    Table 2 from Hoekstra et al. (2014) - + @@ -486,6 +514,9 @@

    Researcher don’t know

    +
    +
    +

    CI compare to H0

    diff --git a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-4-1.png b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-4-1.png index 08325d33..917fb126 100644 Binary files a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-4-1.png and b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-4-1.png differ diff --git a/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-5-1.png b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-5-1.png new file mode 100644 index 00000000..08325d33 Binary files /dev/null and b/courses/SM/2023-2024-S2/alternatives_nhst/alternatives_nhst_files/figure-revealjs/unnamed-chunk-5-1.png differ diff --git a/topics/.DS_Store b/topics/.DS_Store index c5b04286..64e1de9b 100644 Binary files a/topics/.DS_Store and b/topics/.DS_Store differ diff --git a/topics/nhst_reflection/nhst_reflection.qmd b/topics/nhst_reflection/nhst_reflection.qmd new file mode 100644 index 00000000..98919905 --- /dev/null +++ b/topics/nhst_reflection/nhst_reflection.qmd @@ -0,0 +1,23 @@ +# The problem with P-values {background-image="Sad-P.webp" background-color="black"} + +## There is no problem + +The problem with P-values is that they are often **misunderstood** and **misinterpreted**. The P-value is the probability of observing a sample statistic as or more extreme as the one obtained, given that the null hypothesis is true. It is **NOT** the probability that the null hypothesis is true. The P-value is **NOT** a measure of the strength of the evidence against the null hypothesis. + +> The mis interpretation is the problem, +> and not adhering to the Nayman-Pearson framework + +## The dance of the P-value + +* [Should replication reveal the same _p_?](https://youtu.be/ez4DgdurRPg?si=z7oIlKZx6iZjHNYH&t=58) +* [What Power are you using](https://youtu.be/ez4DgdurRPg?si=pN0QTEjARl_2mUO0&t=235) +* [Increasing the power](https://youtu.be/ez4DgdurRPg?si=QQku6BKu4C-8BvhF&t=396) +* [Comparing CI's to single point](https://youtu.be/ez4DgdurRPg?si=QPAcDeFmG-BUe8ZH&t=480) + +## H0 and HA distribution {.center} + +```{r tiny-effects, fig.pos='H', fig.align='center', fig.cap="Any effect can be statistically significant.", echo=FALSE, screenshot.opts = list(delay = 5), dev="png", out.width="1200px"} +# Illustrate that even tiny effects can yield statistically significant test results if the sample is sufficiently large. +# Generate a normal distribution as hypothesized sampling distribution (M = 2.8, SE = SD / sqrt(N) = 0.6 / sqrt(10) = 0.2) with 2.5% of each tail area coloured. Add a vertical line with value for the sample average linked to a slider (range [2.82, 3.00] initial value 2.90). Add a sample size slider (range [10, 5,000], initial value 10), which is linked to the standard error of the normal curve. With slider for (assumed) true population mean and test power. +knitr::include_app("https://sharon-klinkenberg.shinyapps.io/tiny-effects/", height="340px") +``` \ No newline at end of file