---
title: "Father absence analysis Boothroyd Cross 2017"
author: "LB"
date: "19 May 2017"
output:
  word_document: default
---

```{r setup, include=FALSE}
library(knitr)
opts_chunk$set(tidy.opts=list(width.cutoff=60),tidy=TRUE)
```
Load all packages needed in script
```{r}
#clear environment first
rm(list=ls())
#load stuff
library(readr)
library(psych)
library(heplots)
library(ggplot2)
library(tidyr)
library(plyr)
library(dplyr)
library(plotrix)
library(Hmisc)
```

read data file
```{r read data file}
masc_all <- read_delim("LGB masculinity data.txt", "\t", escape_double = FALSE, trim_ws = TRUE)
```


Key variables in the data set are:
 $ country      : chr  "Australia" "Australia" "Australia" "Australia" ...
 $ age          : int  20 20 22 18 19 18 18 18 19 18 ...
 $ gender       : chr  "female" "female" "female" "female" ...
 $ income       : int  NA NA 1 3 2 3 2 2 2 3 ...
 $ wmf2         : int  NA NA NA 6 4 1 1 7 7 4 ...
 $ wmm2         : int  NA NA NA 4 4 9 7 9 9 7 ...
 $ mqbp2        : int  NA NA NA 4 3 2 4 6 3 4 ...
 $ fath_ab      : num  NA NA NA NA NA NA NA NA NA NA ...
 $ f_ab_ever    : int  NA NA NA NA NA NA NA NA NA NA ...
 $ periodyr     : int  NA NA NA 13 12 13 10 13 12 15 ...
 $ sex_yr       : int  NA NA NA 98 16 18 15 13 98 15 ...
 $ fss_total    : num  NA NA NA 1.12 2.49 3.35 NA 1.78 3.94 1.53 ...
 $ fss_a        : num  NA NA NA 1 2.36 2.09 NA 1.82 4.18 1.27 ...
 $ fss_b        : num  NA NA NA 1.06 2.88 4.13 NA 1.25 4.56 1.25 ...
 $ bsri_a       : num  NA NA NA 3.5 2.45 3.35 4.5 5.25 4.9 4.55 ...
 $ bsri_b       : num  NA NA NA 3.8 3.95 5.95 5.05 5.65 6.35 3.8 ...
 $ bsri_c       : num  NA NA NA 4.2 3.95 5.25 4.25 5 4.85 4.1 ...
 $ BSRI_ratio   : num  NA NA NA 0.92 0.62 0.56 0.89 0.93 0.77 1.2 ...
 $ bis          : num  NA NA NA 2.21 1.82 2.03 2.26 2.74 2.47 1.97 ...
 $ bis_attention: num  NA NA NA 2.75 3 2.25 1.25 2 2.38 1.75 ...
 $ bis_motor    : num  NA NA NA 2.7 2.4 2.2 1.7 2.7 2.7 1.8 ...
 $ bis_planning : num  NA NA NA 2.92 2.5 2.17 2.5 1.67 3 2.17 ...
 $ agg          : num  NA NA NA NA 2.19 2.64 2.07 3.24 NA 1.86 ...
 $ agg_ang      : num  NA NA NA NA 2 2 1.57 2.86 NA 1.57 ...
 $ agg_host     : num  NA NA NA NA 1.57 3.75 1.88 3.88 NA 2 ...
 $ agg_phys     : num  NA NA NA NA 2.89 2 1.67 1.78 NA 1.78 ...
 $ agg_verb     : num  NA NA NA NA 2 3.6 3 3.8 NA 1.4 ...
```{r}

#run this to list full list of variables untruncated if needed: 
#str(masc_all, list.len=length(masc_all)) 
```


#Method - Participants Section

First we select participants aged 17 to 30 years and report mean age.
We also recode as NA any reported menarche either before age 9 or after age 17 (i.e. medically precociours/delayed), or reported first sex before age 9 (i.e. before earliest age at puberty)

```{r recode NA ages}
#recode '98' in age first coitus/menarche as NA, for later analyses.
masc_all$sex_yr[masc_all$sex_yr==98] <- NA
masc_all$periodyr[masc_all$periodyr==98] <- NA

#recode values <9 or >17 for age at menarche as NA
#recode values <9 for age at first sex as NA
masc_all$periodyr[masc_all$periodyr<9] <- NA
masc_all$periodyr[masc_all$periodyr>17] <- NA
masc_all$sex_yr[masc_all$sex_yr<9] <- NA

#drop all cases where respondent is under 16, over 30, or did not report age, leaves N = 691 with mean age 20.5 as reported in MS
masc_ageset <- masc_all[is.na(masc_all$age) == FALSE,]
masc_ageset <- masc_ageset[(masc_ageset$age>16), ]
masc_ageset <- masc_ageset[(masc_ageset$age<30), ]
nrow(masc_ageset)

#various frequencies
mean(masc_ageset$age)
table(masc_ageset$white, masc_ageset$country)
table(masc_ageset$country)
table(masc_ageset$gender)
table(masc_ageset$gender, masc_ageset$fath_ab)

```


```{r descriptives}
descriptives <- subset(masc_ageset, select=c(country, gender, white, job, f_ab_ever,	age,	periodyr,	sex_yr,		bsri_a,	bsri_b,	agg_ang,	agg_host,	agg_phys,	agg_verb,fss_total,	bis_attention,	bis_motor,	bis_planning))
descriptives.US <- descriptives[(descriptives$country == "United State"), ]
descriptives.Aus <- descriptives[(descriptives$country == "Australia"), ]

describe(descriptives.US) #gives loads more than needed, but ns etc  all there (from the Hmisc package)
#psych::describe(descriptives.US) #gives sds for continuous variables and a big mess for everything else

describe(descriptives.Aus) #gives loads more than needed, but ns etc  all there (from the Hmisc package)
#psych::describe(descriptives.Aus) #gives sds for continuous variables and a big mess for everything else
```



We're going to pause here and check the factor structure of the FSS

```{r PCA for FSS}
PCAdataF <- subset(masc_ageset, select=c(fss1,	fss2,	fss3,	fss4,	fss5,	fss6,	fss7,	fss8,	fss9,	fss10,	fss11,	fss12,	fss13,	fss14,	fss15,	fss16,	fss17,	fss18,	fss19,	fss20,	fss21,	fss22,	fss23,	fss24,	fss25,	fss26,	fss27,	fss28,	fss29,	fss30,	fss31,	fss32,	fss33,	fss34,	fss35,	fss36,	fss37,	fss38,	fss39,	fss40,	fss41,	fss42,	fss43,	fss44,	fss45,	fss46,	fss47,	fss48,	fss49,	fss50,	fss51,	fss52,	fss53,	fss54,	fss55,	fss56,		fss57a,	fss57b,	fss58,	fss59,	fss60,	fss61,	fss62,	fss63,	fss64,	fss65,	fss66,	fss67,	fss68,	fss69,	fss70a,	fss70b,	fss71,	fss72,	fss73,	fss74,	fss75,	fss76,	fss77,	fss78,	fss79,	fss80,	fss81,	fss82,	fss83,	fss84,	fss85,	fss86,	fss87))
#scale(PCAdataF, center = TRUE, scale = TRUE)

library(psych)
fitF <- principal(PCAdataF, nfactors=4, rotate="none")
print(fitF,cut=0.3,digits=3) # print results
scree(PCAdataF)

#it looks like fss_total is the best variable to use

```


Then we check whether age is associated with any of our individual outcome variables.

```{r Adjusting for age}
#does age predict any of our outcome variables?
#a couple of sig parameter estimates within nonsig models, or nonsig parameter estimates within sig models, but nothing in here to suggest adjusting for age is necessary 

#make an agesquared variable to check for quadratic effects
masc_ageset$agesq <- masc_ageset$age*masc_ageset$age

#does age predict fear a?
# note na.action = na.exclude - this pads out the residuals with NAs in the appropriate places rather than just producing nothing. Lets you bind the residuals to the data frame afterwards.

#FSS_total has sig model but all driven by intercept
fss_total.age <- lm(fss_total ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(fss_total.age) 

#not age at menarche
periodyr.age <- lm(periodyr ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(periodyr.age)

#not BSRI masc
bem_a.age <- lm(bsri_a ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(bem_a.age)

#not BSRI fem
bem_b.age <- lm(bsri_b ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(bem_b.age)

#not AQ anger
ang.age <- lm(agg_ang ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(ang.age)

#not AQ hostile
host.age <- lm(agg_host ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(host.age)

#AQ phys - weak variable coeffs, nonsig model
phys.age <- lm(agg_phys ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(phys.age)

#AQ verbs - weak variable coeffs, nonsig model
verb.age <- lm(agg_verb ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(verb.age)

#not BIS motor 
BISmot.age <- lm(bis_motor ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(BISmot.age)

#not BIS attention
BISatt.age <- lm(bis_attention ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(BISatt.age)

#not BIS planning
BISplan.age <- lm(bis_planning ~ age + agesq, data = masc_ageset, na.action = na.exclude)
summary(BISplan.age)

```

We also need to do the PCA on family ratings.
```{r}

PCAdata1 <- subset(masc_ageset, select=c(wmf2, wmm2, mqbp2))
#scale(PCAdata1, center = TRUE, scale = TRUE)

fit1 <- principal(PCAdata1, rotate="none")
print(fit1,cut=0.3,digits=3) # print results


# add factor scores to dataset
masc_ageset$FRQ <- fit1$scores[,1]
```

Now we do the big PCA on all gendered behaviour items

```{r}

PCAdata2 <- subset(masc_ageset, select=c(bsri_a,bsri_b, agg_ang, agg_host, agg_phys, agg_verb,fss_total, bis_attention, bis_motor, bis_planning))
#scale(PCAdata2, center = TRUE, scale = TRUE)

library(psych)
fit2 <- principal(PCAdata2, nfactors=4, rotate="none")
print(fit2,cut=0.3,digits=3) # print results
scree(PCAdata2)


# add factor scores to dataset
masc_ageset$reactivity <- fit2$scores[,1]
masc_ageset$masculinity <- fit2$scores[,2]
```

Now we check for sex differences on these factors. We find that reactivity shows a country difference (Aus scores higher) but no sex difference. Masculinity shows a sex diference (men score higher) and a country difference (US scores higher) but no interaction.

```{r gender and country differences in factor scores}
r_sexdiff <- lm(reactivity ~ gender*country, data = masc_ageset, na.action = na.exclude)
summary(r_sexdiff) 
etasq(r_sexdiff)
m_sexdiff <- lm(masculinity ~ gender*country, data = masc_ageset, na.action = na.exclude)
summary(m_sexdiff)  
etasq(m_sexdiff)


#I like Cohen's d becuase I find it intuitive. So I've made it here. Yes, I know this code is *horrible*

#drop cases with NA factor scores
for_describing_factors <- masc_ageset[is.na(masc_ageset$masculinity) == FALSE,]
for_describing_factors <- for_describing_factors[is.na(for_describing_factors$reactivity) == FALSE,]
#spit out means for M and then F
ddply(for_describing_factors, .(gender), summarise, mascmean=mean(masculinity))
ddply(for_describing_factors, .(gender), summarise, reacmean=mean(reactivity))
#spit out SDs for M and then F
ddply(for_describing_factors, .(gender), summarise, mascSD=sd(masculinity))
ddply(for_describing_factors, .(gender), summarise, reacSD=sd(reactivity))
#spit out sd for (M+F)
sd(for_describing_factors$reactivity)
sd(for_describing_factors$masculinity)
#from this, Cohen's d for sdiff in masc is (0.27653171 - (-0.07570487))/0.9693821 = 0.36
#Cohen's d for sdiff in reac is (0.03943915 - (-0.02470656))/0.9942785 = 0.06


#look at the factor scores by country
#drop cases with NA factor scores
for_describing_factors_country <- masc_ageset[is.na(masc_ageset$masculinity) == FALSE,]
for_describing_factors_country <- for_describing_factors_country[is.na(for_describing_factors_country$reactivity) == FALSE,]
#spit out means for M and then F
ddply(for_describing_factors_country, .(country), summarise, reacmean=mean(reactivity))
ddply(for_describing_factors_country, .(country), summarise, mascmean=mean(masculinity))
#spit out separeate SDs for M and F
ddply(for_describing_factors_country, .(country), summarise, reacSD=sd(reactivity))
ddply(for_describing_factors_country, .(country), summarise, mascSD=sd(masculinity))
#spit out sd for (M+F)
sd(for_describing_factors_country$reactivity)
sd(for_describing_factors_country$masculinity)
#from this, Cohen's d for country diff in reac is (0.09236556 - (-0.30040209))/0.9942785 = 0.3950279
#Cohen's d for country diff in masc is (0.2426770 - (-0.1073859))/0.9693821 = 0.3611196


```


#father absence analysis

We select US women and men and Aussie men for father absence analyses (becuase N too small for US males), and drop anyone who is father absent NA or blank

```{r}
masc <- masc_ageset[masc_ageset$country == "Australia" | (masc_ageset$gender == "female" & masc_ageset$country == "United State"), ]
nrow(masc)
table(masc$country, masc$gender)

## Remove anyone with fath_ab of 98 (n=45) or NA (n=56). Leaves total n=654
table(masc$fath_ab, useNA = "always")
masc <- masc[(masc$fath_ab == 1 | masc$fath_ab == 0.5 | masc$fath_ab == 0) & is.na(masc$fath_ab) == FALSE,]
nrow(masc)
```


Now we run father absence analyses on this subset

3-category father absence for reactivity in female participants: shows clear effect of country, with Americans less reactive than Australians. Effects of fath_abnon-sig; remain non-sig if income added to model
```{r}
masc$fath_ab <- as.factor(masc$fath_ab)  # make sure fath_ab is a factor not continuous
father.absence.reactivityf <- lm(reactivity ~ fath_ab * country , data = masc[masc$gender == "female",])
summary(father.absence.reactivityf)
etasq(father.absence.reactivityf)

#Here's the model with income too
#father.absence.reactivityf.income <- lm(reactivity ~ fath_ab * country + income , data = masc[masc$gender == "female",])
#summary(father.absence.reactivityf.income)
#etasq(father.absence.reactivityf.income)

```


3-category father absence for masculinity in female participants: Weak non-sig effect of fath_ab=1 vs fath_ab=0, US clearly less masculine, effect of fath_ab 1 on masculinity weaker in US than AUS. Adding income doesn't change any of these
```{r}
masc$fath_ab <- as.factor(masc$fath_ab)  # make sure fath_ab is a factor not continuous
father.absence.masculinityf <- lm(masculinity ~ fath_ab * country, data = masc[masc$gender == "female",])
summary(father.absence.masculinityf)
etasq(father.absence.masculinityf)

#Here's the model with income too
#father.absence.masculinityf.income <- lm(masculinity ~ fath_ab * country + income, data = masc[masc$gender == "female",])
#summary(father.absence.masculinityf.income)
#etasq(father.absence.masculinityf.income)
```
We detour to investigate that interaction. There's a weak, non-sig effect of Fath_ab=1 vs 0 in Australia, nothing at all in America. Income makes no difference.

```{r}
father.absence.masculinity.Aus <- lm(masculinity ~ fath_ab , data = masc[(masc$gender == "female")&(masc$country=="Australia"),])
summary(father.absence.masculinity.Aus)
etasq(father.absence.masculinity.Aus)

#Here's the model with income too
#father.absence.masculinity.Aus.income <- lm(masculinity ~ fath_ab +income, data = masc[(masc$gender == "female")&(masc$country=="Australia"),])
#summary(father.absence.masculinity.Aus.income)
#etasq(father.absence.masculinity.Aus.income)

father.absence.masculinity.US <- lm(masculinity ~ fath_ab , data = masc[(masc$gender == "female")&(masc$country=="United State"),])
summary(father.absence.masculinity.US)
etasq(father.absence.masculinity.US)

#Here's the model with income too
#father.absence.masculinity.US.income <- lm(masculinity ~ fath_ab +income, data = masc[(masc$gender == "female")&(masc$country=="United State"),])
#summary(father.absence.masculinity.US.income)
#etasq(father.absence.masculinity.US.income)

```


2-category father absence for reactivity in female participants: strong effect of country again; weak but sig  effect of father absence 
```{r}
masc$f_ab_ever <- as.factor(masc$f_ab_ever)  # make sure fath_ab is a factor not continuous
father.absence.reactivity2f <- lm(reactivity ~ f_ab_ever*country, data = masc[masc$gender == "female",])
summary(father.absence.reactivity2f)
etasq(father.absence.reactivity2f)

#Here's the model with income too
#father.absence.reactivity2f.income <- lm(reactivity ~ f_ab_ever*country+income, data = masc[masc$gender == "female",])
#summary(father.absence.reactivity2f.income)
#etasq(father.absence.reactivity2f.income)
```


2-category father absence for masculinity in female participants: no effects of FA, again Americans are less masculine
```{r}
masc$f_ab_ever <- as.factor(masc$f_ab_ever)  # make sure fath_ab is a factor not continuous
father.absence.masculinity2f <- lm(masculinity ~ f_ab_ever*country, data = masc[masc$gender == "female",])
summary(father.absence.masculinity2f)
etasq(father.absence.masculinity2f)

#Here's the model with income too
#father.absence.masculinity2f.income <- lm(masculinity ~ f_ab_ever*country+income, data = masc[masc$gender == "female",])
#summary(father.absence.masculinity2f.income)
#etasq(father.absence.masculinity2f.income)
```


2-category father absence for reactivity in male AUS participants: no effects
```{r}
masc$f_ab_ever <- as.factor(masc$f_ab_ever)  # make sure fath_ab is a factor not continuous
father.absence.reactivity2m <- lm(reactivity ~ f_ab_ever, data = masc[masc$gender == "male",])
summary(father.absence.reactivity2m)
etasq(father.absence.reactivity2m)

#Here's the model with income too
#father.absence.reactivity2m.income <- lm(reactivity ~ f_ab_ever+income, data = masc[masc$gender == "male",])
#summary(father.absence.reactivity2m.income)
#etasq(father.absence.reactivity2m.income)
```


2-category father absence for masculinity in male AUS participants: no effect of father absence 
```{r}
masc$f_ab_ever <- as.factor(masc$f_ab_ever)  # make sure fath_ab is a factor not continuous
father.absence.masculinity2m <- lm(masculinity ~ f_ab_ever, data = masc[masc$gender == "male",])
summary(father.absence.masculinity2m)
etasq(father.absence.masculinity2m)

#Here's the model with income too
#father.absence.masculinity2m.income <- lm(masculinity ~ f_ab_ever+income, data = masc[masc$gender == "male",])
#summary(father.absence.masculinity2m.income)
#etasq(father.absence.masculinity2m.income)
```

#Using relationship quality to predict factor scores.

Does quality predict reactivity in women?
yes - better relationship -> lower reactivity
also effect of country (lower reactivity in US than Aus)
no interaction

```{r}
quality.reactivity.f <- lm(reactivity ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "female",])
etasq(quality.reactivity.f)
summary(quality.reactivity.f)

#Here's the model with income too
#quality.reactivity.f.income <- lm(reactivity ~ FRQ*country+income, data = masc_ageset[masc_ageset$gender == "female",])
#etasq(quality.reactivity.f.income)
#summary(quality.reactivity.f.income)
```

Does quality predict masculinity in women?
No!
sig effect of country, no interaction
R squared 0.03
```{r}
quality.masculinity.f <- lm(masculinity ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "female",])
summary(quality.masculinity.f)
etasq(quality.masculinity.f)

#Here's the model with income too
#quality.masculinity.f.income <- lm(masculinity ~ FRQ*country+income, data = masc_ageset[masc_ageset$gender == "female",])
#summary(quality.masculinity.f.income)
#etasq(quality.masculinity.f.income)
```
Does quality predict reactivity in men?
yes - better relationship -> lower reactivity
also effect of country (lower reactivity in US than Aus)
no interaction

```{r}
quality.reactivity.m <- lm(reactivity ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "male",])
etasq(quality.reactivity.m)
summary(quality.reactivity.m)

#Here's the model with income too
#quality.reactivity.m.income <- lm(reactivity ~ FRQ*country+income, data = masc_ageset[masc_ageset$gender == "male",])
#etasq(quality.reactivity.m.income)
#summary(quality.reactivity.m.income)
```
Does quality predict masculinity in men?
No!
marginal effect of country, no interaction
R squared 0.03

```{r}
quality.masculinity.m <- lm(masculinity ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "male",])
summary(quality.masculinity.m)
etasq(quality.masculinity.m)

#Here's the model with income too
#quality.masculinity.m.income <- lm(masculinity ~ FRQ*country+income, data = masc_ageset[masc_ageset$gender == "male",])
#summary(quality.masculinity.m.income)
#etasq(quality.masculinity.m.income)
```


#Finally we check how everything relates to reproductive outcomes.


 is father absence related to age at first coitus in women?
lm predicting first coitus age with country, father-absence, and interaction
3-category father absence analysis: 
sig interaction between country and fath-ab1
```{r}
masc$fath_ab <- as.factor(masc$fath_ab)  # make sure fath_ab is a factor not continuous
father.absence.sex.f <- lm(sex_yr ~ fath_ab*country, data = masc[masc$gender == "female",])
summary(father.absence.sex.f)
etasq(father.absence.sex.f)

#income added to model
#father.absence.sex.f.income <- lm(sex_yr ~ fath_ab*country + income, data = masc[masc$gender == "female",])
#summary(father.absence.sex.f.income)
#etasq(father.absence.sex.f.income)
```
Looking at that interaction for coitus in women, there's a significant impact of father absence pre-12 in the US (those father absent when young have sex ~1.5 years earlier than father present). Nothing in Australia


```{r}
father.absence.sex.Aus <- lm(sex_yr ~ fath_ab , data = masc[(masc$gender == "female")&(masc$country=="Australia"),])
summary(father.absence.sex.Aus)
etasq(father.absence.sex.Aus)

#Here's the model with income too
#father.absence.sex.Aus.income <- lm(sex_yr ~ fath_ab +income, data = masc[(masc$gender == "female")&(masc$country=="Australia"),])
#summary(father.absence.sex.Aus.income)
#etasq(father.absence.sex.Aus.income)

father.absence.sex.US <- lm(sex_yr ~ fath_ab , data = masc[(masc$gender == "female")&(masc$country=="United State"),])
summary(father.absence.sex.US)
etasq(father.absence.sex.US)

#Here's the model with income too
#father.absence.sex.US.income <- lm(sex_yr ~ fath_ab +income, data = masc[(masc$gender == "female")&(masc$country=="United State"),])
#summary(father.absence.sex.US.income)
#etasq(father.absence.sex.US.income)
```


is father absence related to age at first coitus in (Aus) men?
No
```{r}
masc$fath_ab <- as.factor(masc$fath_ab)  # make sure fath_ab is a factor not continuous
father.absence.sex.m <- lm(sex_yr ~ f_ab_ever, data = masc[masc$gender == "male",])
summary(father.absence.sex.m)
etasq(father.absence.sex.m)

#income added to model
#father.absence.sex.m.income <- lm(sex_yr ~ f_ab_ever + income, data = masc[masc$gender == "male",])
#summary(father.absence.sex.m.income)
#etasq(father.absence.sex.m.income)
```


 is father absence related to age at menarche in women?
no clear evidence of effect of fath-ab, effect of country, or interaction
```{r}
masc$fath_ab <- as.factor(masc$fath_ab)  # make sure fath_ab is a factor not continuous
father.absence.menarche <- lm(periodyr ~ fath_ab*country, data = masc)
summary(father.absence.menarche)
etasq(father.absence.menarche)

#income added to model
father.absence.menarche.income <- lm(periodyr ~ fath_ab*country + income, data = masc)
summary(father.absence.menarche.income)
etasq(father.absence.menarche.income)
```



does quality of family relationships predict reproductive outcomes?
NB notice we've shifted back to the original age-selected data set as missing father absence data not a problem here and plenty of US males.
first: does it predict age at menarche?
no 

```{r}
quality.menarche <- lm(periodyr ~ FRQ*country, data = masc_ageset)
summary(quality.menarche)
etasq(quality.menarche)

#income added to model
#quality.menarche.income <- lm(periodyr ~ FRQ*country + income, data = masc_ageset)
#summary(quality.menarche.income)
#etasq(quality.menarche.income)
```


does quality of family relationships predict age at first coitus in women?
sig interaction with country

```{r}
quality.fc.f <- lm(sex_yr ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "female",])
summary(quality.fc.f)
etasq(quality.fc.f)

#Model with income added
#quality.fc.f.income <- lm(sex_yr ~ FRQ*country+income, data = masc_ageset[masc_ageset$gender == "female",])
#summary(quality.fc.f.income)
#etasq(quality.fc.f.income)
```
that interaction for coitus and relationship quality in women, there's again a  significant association in the US which remains when income is added to the model. Nothing in Australia either way.


```{r}
#Aus women 
quality.sex.Aus <- lm(sex_yr ~ FRQ , data = masc_ageset[(masc_ageset$gender == "female")&(masc_ageset$country=="Australia"),])
summary(quality.sex.Aus)
etasq(quality.sex.Aus)
#Here's the model with income too
#quality.sex.Aus.income <- lm(sex_yr ~ FRQ +income, data = masc_ageset[(masc_ageset$gender == "female")&(masc_ageset$country=="Australia"),])
#summary(quality.sex.Aus.income)

#US women
quality.sex.US <- lm(sex_yr ~ FRQ , data = masc_ageset[(masc_ageset$gender == "female")&(masc_ageset$country=="United State"),])
summary(quality.sex.US)
etasq(quality.sex.US)
#Here's the model with income too
#quality.sex.US.income <- lm(sex_yr ~ FRQ +income, data = masc_ageset[(masc_ageset$gender == "female")&(masc_ageset$country=="United State"),])
#summary(quality.sex.US.income)

```
does quality of family relationships predict age at first coitus in men?
no

```{r}

quality.fc.m <- lm(sex_yr ~ FRQ*country, data = masc_ageset[masc_ageset$gender == "male",])
summary(quality.fc.m)
etasq(quality.fc.m)

#checking model with income
#quality.fc.m.income <- lm(sex_yr ~ FRQ*country + income, data = masc_ageset[masc_ageset$gender == "male",])
#summary(quality.fc.m.income)
#etasq(quality.fc.m.income)
```



#Checking income as a predictor of our measures of interest
is father absence associated with lower income (regression other way round bc factors)
```{r}
#not (quite) age at 1st sex
income.sex_yr <- lm(sex_yr ~ income, data = masc_ageset)
summary(income.sex_yr) 

#not age at menarche
income.periodyr <- lm(periodyr ~ income, data = masc_ageset)
summary(income.periodyr)

#not reactivity
income.reactivity <- lm(reactivity ~ income, data = masc_ageset)
summary(income.reactivity)

#not masculinity
income.masculinity <- lm(masculinity ~ income, data = masc_ageset)
summary(income.masculinity)

#poorer relationships, associated with lower income
income.quality <- lm(FRQ ~ income, data = masc_ageset)
summary(income.quality)
etasq(income.quality)

#father absence
father.absence.income <- lm(income ~ fath_ab, data = masc)
summary(father.absence.income)
etasq(father.absence.income)

```

graph age of first coitus in women
```{r}
#subset relevant variables
fcfgraph <- subset(masc, select=c(fath_ab, country, gender, sex_yr))
#women only
fcfgraph <-  fcfgraph[fcfgraph$gender == "female",]
#drop na values
fcfgraph <- fcfgraph %>% drop_na()
nrow(fcfgraph)

fcfgraph$fath_ab <- as.factor(fcfgraph$fath_ab)
fcfgraph$country <- as.factor(fcfgraph$country)
fcfgraph <- as.data.frame(fcfgraph)

fcf_means <- ddply(fcfgraph, c("fath_ab"), summarise,
                    mean=mean(sex_yr))
fcf_ses <- ddply(fcfgraph, c("fath_ab"), summarise,
                  se=std.error(sex_yr))


#add the se column to the dataframe with means in it and check that the new dataframe is OK
#yes, I know
fcf_means$se <- fcf_ses$se
str(fcf_means)


#make bargraph of age at 1st coitus score +/- 2ses by f_ab_ever
gfcf <- ggplot(data = fcf_means, 
                 aes(x=fath_ab,
                     y= mean, 
                     ymin=mean-(2*se), 
                     ymax=mean+(2*se))) +
  geom_bar(position="dodge", stat = "identity") + 
  geom_errorbar(position = position_dodge(0.9), width = 0.2, colour="black") +
  xlab("Father absent") +
  ylab("Age at first coitus") +
  scale_x_discrete(labels=c("0" = "Never", "0.5" = "After age 12", "1" = "Before age 12")) +
  theme_classic()
gfcf
```