---
tags: reviews
---

# Review: Effects of variables on review grading

### Updated 20/1 2022: analysis done on the goldstandard dataset.

### TODO: standardize the grade (d-mean)

### TODO: Look at the author gender

## Gradings in newspapers, blogs, and online media:

#### Mean and std:


### Grade distribution


## Grades given female and male reviewers




So we can see that female and male reviewers and grading very similarly. But we can also see that the grading given in blogs and newspapers differ.
How is gender distributed among media types?

## Gender of reviewer in different media type


## Effects of reviewer gender on review grade (OLS)

By making a linear regression (ordinary linear-squared), we can see how variables affect the outcome.

First I look at the grade (grades_transformed_6) as the outcome and the gender of the reviewer (rev_gender_updated) as the predicter.

Here we see a negative coefficient for male reviewers ==> male reviewers are giving 0.1869 lower grades compares to female reviewers. Beeing statistically significant with the t-value < 1.96.


## Effects of media type on review grade (OLS)

Now we look at the grade (grades_transformed_6) as the outcome and the media type (media_name_type) as the predicter.

Here we can see that lowest grade is given in newspapers (being the baseline shown as Intercept). Highest grades are given in 'Blogs'.
All findings except those for 'Regionalavis' are statistically significant with |t|> 1.96


## Interactive effect of gender and media type (OLS)

How is gender and media type interacting?

**Female reviewers:**

* For **female reviewers in newspapers: 4.0092**
* For female reviewers in blogs: 4.0092 + 0.5274 = **4.5366**
* For female reviewers in Fagblad: _not statistically significant_
* For **female reviewers in Online Media**: 4.0092 + 0.3194 = **4.3286**
* For **female reviewers in Online Media without quote**: 4.0092 + 0.5023 = **4.5115**
* For **female reviewers in Regional avis**: 4.0092 + 0.1481 = = **4.1573**
* For **female reviewers in Ugeblad**: 4.0092 + 0.5058 = **4.5150**

**Male reviewers:**
* For **male reviewers in newspapers:** 4.0092 + 0.1562 = **4.1654**
* For **male reviewers in Blog**s: 4.0092 + 0.1562 - 0.3579 = **3.8075**
* For **male reviewers in Fagblad**: 4.0092 + 0.1562 + 0.4876 = **4.653**
* For **male reviewers in Online Media**: 4.0092 + 0.1562 - 0.3279 = **3.8375**
* For male reviewers in Online Media without quote: _not statistically significant_
* For **male reviewers in Regional avis**: 4.0092 + 0.1562 - 0.2043 = **3.9611**
* For male reviewers in Ugeblad: _non existing_

Female reviewers are "tougher" in newspapers than men. But _way_ "nicer" in blogs and online media (with and without quoute) than men.
Male reviewers gives the lowest grades in the media types where female reviewers are giving the highest grade.

Keeping in mind the distribution of reviewer gender in different media types, the dominating reviewer gender gives the highest grade (male in newspapers and regional newspapers, female in blogs and online media)