| Variable | Summary |
|---|---|
| PK1 [correct], % | 83.1 |
| PK2 [correct], % | 68.7 |
| PK3 [correct], % | 40.7 |
| PK4 [correct], % | 87.9 |
| PK5 [correct], % | 62.8 |
| PK6 [correct], % | 24.1 |
Methoden der empirischen Kommunikations- und Medienforschung
Freie Universität Berlin
Freitag, 3.7., 10 UHR, Online Q&A zur Klausur (Ersatz für 8. 6.)
Bitte an Selbstreport der Hausaufgaben denken — Voraussetzung für Bestätigung der aktiven Teilnahme
→ Funktion zwischen Prädiktoren und aV ist nicht unbedingt linear
→ Lineares Modell sagt Werte voraus, die es nicht geben kann
| Variable | Summary |
|---|---|
| PK1 [correct], % | 83.1 |
| PK2 [correct], % | 68.7 |
| PK3 [correct], % | 40.7 |
| PK4 [correct], % | 87.9 |
| PK5 [correct], % | 62.8 |
| PK6 [correct], % | 24.1 |

| id | likes_count | comments_count | shares_count |
|---|---|---|---|
| 18058830773_315819235204024 | 105 | 9 | 10 |
| 140105122708_10152295975152709 | 16 | 3 | 2 |
| 103256838688_10153391508503689 | 330 | 10 | 30 |
| 49702985881_10151656671260882 | 10 | 0 | 3 |
| 16761458703_10153232709798704 | 161 | 1 | 6 |
| 374809010319_10151487539830320 | 1331 | 35 | 204 |
| 18058830773_10151884419680774 | 62 | 2 | 7 |
| 374809010319_10152497007995320 | 49 | 1 | 4 |
| 55528788113_10151241262778114 | 390 | 17 | 21 |
| 58323112191_10151982898231845 | 7 | 0 | 0 |
Simuliertes Beispiel!

Simuliertes Beispiel!

\(Logit(Y_i) = b_0 + b_1 X_i\)


| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | -0.38 | (-0.50, -0.25) | -5.84 | < .001 | |
| Tjur’s R2 | 0 |
\(Logit(Y) = b_0\)
\(P(Y) = \frac {1}{1+e^{-(\beta_0)}}\)



| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | -0.39 | (-0.52, -0.26) | -5.97 | < .001 | |
| Age10 c | 0.28 | (0.18, 0.38) | 5.60 | < .001 | |
| Tjur’s R2 | 0.03 |

Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.066 0.011 5.9 <0.001 28.2 0.044 0.088
Term: Age10_c
Type: response
Comparison: +1
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
-0.063 0.01 -6.1 <0.001 30.2 -0.083 -0.043
Term: Age10_c
Type: response
Comparison: +-1
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.049 0.0052 9.6 <0.001 69.9 0.039 0.06
Term: Age10_c
Type: response
Comparison: -2.397583081571 - -3.397583081571
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.069 0.012 5.6 <0.001 25.4 0.045 0.094
Term: Age10_c
Type: response
Comparison: 1.802416918429 - 0.802416918429003
| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | 0.67 | (0.59, 0.77) | -5.97 | < .001 | |
| Age10 c | 1.32 | (1.20, 1.46) | 5.60 | < .001 | |
| Tjur’s R2 | 0.03 |
d <- d |>
mutate(
Online_news_sites_c = Online_news_sites - mean(Online_news_sites),
Twitter_c = Twitter - mean(Twitter),
Facebook_c = Facebook - mean(Facebook),
Age10_c = Age10 - mean(Age10)
)
m2 <- glm(
PK3 ~ Online_news_sites_c + Twitter_c + Facebook_c + Gender + Age10_c, # Regressionsgleichung
family = binomial(link = "logit"), # aV ist binär, logit-Link = Logistische Regression
data = d
)| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | 0.13 | (-0.05, 0.31) | 1.42 | 0.157 | |
| Online news sites c | 0.19 | (0.11, 0.28) | 4.50 | < .001 | |
| Twitter c | -0.03 | (-0.18, 0.12) | -0.42 | 0.675 | |
| Facebook c | -0.11 | (-0.19, -0.04) | -2.82 | 0.005 | |
| Gender (female) | -1.20 | (-1.48, -0.92) | -8.35 | < .001 | |
| Age10 c | 0.19 | (0.09, 0.30) | 3.55 | < .001 | |
| Tjur’s R2 | 0.14 |
Term Contrast Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Age10_c +1 0.0406 0.0113 3.60 <0.001 11.6 0.019 0.0627
Facebook_c +1 -0.0235 0.0081 -2.89 0.0039 8.0 -0.039 -0.0076
Gender female - male -0.2664 0.0306 -8.69 <0.001 58.0 -0.326 -0.2063
Online_news_sites_c +1 0.0406 0.0088 4.63 <0.001 18.0 0.023 0.0578
Twitter_c +1 -0.0067 0.0159 -0.42 0.6745 0.6 -0.038 0.0245
Type: response
| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | 1.14 | (0.95, 1.36) | 1.42 | 0.157 | |
| Online news sites c | 1.21 | (1.12, 1.32) | 4.50 | < .001 | |
| Twitter c | 0.97 | (0.83, 1.12) | -0.42 | 0.675 | |
| Facebook c | 0.89 | (0.82, 0.97) | -2.82 | 0.005 | |
| Gender (female) | 0.30 | (0.23, 0.40) | -8.35 | < .001 | |
| Age10 c | 1.21 | (1.09, 1.35) | 3.55 | < .001 | |
| Tjur’s R2 | 0.14 |
| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | 0.13 | (-0.05, 0.31) | 1.42 | 0.157 | |
| Online news sites c | 0.19 | (0.11, 0.28) | 4.50 | < .001 | |
| Twitter c | -0.03 | (-0.18, 0.12) | -0.42 | 0.675 | |
| Facebook c | -0.11 | (-0.19, -0.04) | -2.82 | 0.005 | |
| Gender (female) | -1.20 | (-1.48, -0.92) | -8.35 | < .001 | |
| Age10 c | 0.19 | (0.09, 0.30) | 3.55 | < .001 | |
| Tjur’s R2 | 0.14 |
The model’s explanatory power is moderate (Tjur’s R2 = 0.14)
| Variable | Summary |
|---|---|
| PK1 [correct], % | 83.1 |
| PK2 [correct], % | 68.7 |
| PK3 [correct], % | 40.7 |
| PK4 [correct], % | 87.9 |
| PK5 [correct], % | 62.8 |
| PK6 [correct], % | 24.1 |

m5q_v1 <- glm(
cbind(Political_knowledge, 5 - Political_knowledge) ~ Age10_c, # Zahl der richtig und falsch beantworteten Fragen in cbind()
family = binomial(link = "logit"), # eine Frage ist binär, logit-Link = Logistische Regression
data = d
)
m5q_v2 <- glm(
Political_knowledge / 5 ~ Age10_c, # Anteil der richtig beantworteten Fragen als aV
family = binomial(link = "logit"), # eine Frage ist binär, logit-Link = Logistische Regression
weights = rep(5, nrow(d)), # Zahl der Fragen als Gewicht
data = d
)| Parameter | Coefficient | 95% CI | z | p | Fit |
|---|---|---|---|---|---|
| (Intercept) | 0.45 | (0.40, 0.51) | 15.34 | < .001 | |
| Age10 c | 0.25 | (0.21, 0.29) | 11.78 | < .001 | |


| M | Var | n |
|---|---|---|
| 3.55 | 28.85 | 128 |
Simuliertes Beispiel!

Simuliertes Beispiel!

\(\log(Y_i) = b_0 + b_1 X_i\)
\(Y_i = e^{b_0 + b_1 X_i}\)


| Parameter | Coefficient | 95% CI | z | df | p | Fit |
|---|---|---|---|---|---|---|
| (Intercept) | 1.27 | (1.02, 1.52) | 10.06 | Inf | < .001 | |
| R2_Nagelkerke | 0.00 |



| Parameter | Coefficient | 95% CI | z | df | p | Fit |
|---|---|---|---|---|---|---|
| (Intercept) | 3.22 | (2.57, 4.07) | 9.97 | Inf | < .001 | |
| ym int half | 1.05 | (1.03, 1.07) | 4.20 | Inf | < .001 | |
| R2_Nagelkerke | 0.21 |

Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.15 0.041 3.7 <0.001 12.3 0.073 0.24
Term: ym_int_half
Type: response
Comparison: 1 - 0
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.15 0.038 3.9 <0.001 13.1 0.073 0.22
Term: ym_int_half
Type: response
Comparison: 0 - -1
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.066 0.0082 8.1 <0.001 50.1 0.05 0.083
Term: ym_int_half
Type: response
Comparison: -17 - -18
Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
0.33 0.14 2.3 0.022 5.5 0.047 0.61
Term: ym_int_half
Type: response
Comparison: 17 - 16
| Parameter | Coefficient | 95% CI | z | df | p | Fit |
|---|---|---|---|---|---|---|
| (Intercept) | 2.54 | (1.82, 3.61) | 5.31 | Inf | < .001 | |
| ym int half | 1.05 | (1.03, 1.08) | 4.40 | Inf | < .001 | |
| topic research (yes) | 1.23 | (0.77, 1.96) | 0.89 | Inf | 0.376 | |
| topic teaching (yes) | 1.81 | (1.00, 3.45) | 1.92 | Inf | 0.055 | |
| word count 100 c | 1.12 | (0.31, 4.29) | 0.19 | Inf | 0.848 | |
| R2_Nagelkerke | 0.26 |
Term Contrast Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
topic_research yes - no 0.68 0.776 0.88 0.38 1.4 -0.838 2.20
topic_teaching yes - no 2.52 1.644 1.53 0.13 3.0 -0.706 5.74
word_count_100_c +1 0.39 2.141 0.18 0.86 0.2 -3.809 4.58
ym_int_half 1 - 0 0.16 0.041 3.82 <0.001 12.9 0.076 0.24
Type: response
| Parameter | Coefficient | 95% CI | z | df | p | Fit |
|---|---|---|---|---|---|---|
| (Intercept) | 2.54 | (1.82, 3.61) | 5.31 | Inf | < .001 | |
| ym int half | 1.05 | (1.03, 1.08) | 4.40 | Inf | < .001 | |
| topic research (yes) | 1.23 | (0.77, 1.96) | 0.89 | Inf | 0.376 | |
| topic teaching (yes) | 1.81 | (1.00, 3.45) | 1.92 | Inf | 0.055 | |
| word count 100 c | 1.12 | (0.31, 4.29) | 0.19 | Inf | 0.848 | |
| R2_Nagelkerke | 0.26 |
The model’s explanatory power is substantial (Nagelkerke’s R2 = 0.26)
| Parameter | Poisson | Negativ-binomial |
|---|---|---|
| (Intercept) | 2.46 (2.10, 2.88) | 2.54 (1.80, 3.59) |
| ym int half | 1.06 (1.05, 1.07) | 1.05 (1.03, 1.08) |
| topic research (yes) | 1.31 (1.08, 1.58) | 1.23 (0.78, 1.96) |
| topic teaching (yes) | 1.72 (1.37, 2.17) | 1.81 (0.99, 3.33) |
| word count 100 c | 1.10 (0.66, 1.83) | 1.12 (0.36, 3.48) |
| Observations | 128 | 128 |
# Overdispersion test
dispersion ratio = 5.822
Pearson's Chi-Squared = 716.094
p-value = < 0.001
# Overdispersion test
dispersion ratio = 0.933
p-value = 0.976
| Name | Model | AIC (weights) | BIC (weights) |
|---|---|---|---|
| pm2 | glm | 876.9 (<.001) | 891.1 (<.001) |
| nbm2 | negbin | 589.2 (>.999) | 606.3 (>.999) |
| #Df | LogLik | Df | Chisq | Pr(>Chisq) |
|---|---|---|---|---|
| 5 | -433.44 | |||
| 6 | -288.59 | 1 | 289.70 | 0.00 |
Wiederholung und Fragen
Marko Bachl