Exploring the interdependence of couples’ rest-wake cycles: an actigraphic
Robert Meadows1, Sara Arber1, Susan Venn1, Jenny Hislop2 and Neil Stanley3
1 Department of Sociology, University of Surrey, Guildford, Surrey GU2 7XH, UK
2 Institute of Life Course Studies, Keele University, Keele, Staffs ST5 5BG, UK
3 Clinical Research and Trials Unit, Norfolk and Norwich University Hospital, UK
This study was funded by UK Economic and Social Research Council grant RES-000-23-
Centre for Research on Ageing and Gender
Department of Sociology
School of Human Sciences
University of Surrey
Tel: 01483 689292
Fax: 01483 689090
Within western societies it is commonplace for couples to share a bed. Yet there has been
remarkably little research carried out on couples’ sleep. This paper draws upon actigraphy,
audio diary and questionnaire data from both partners in 36 heterosexual couples (age 20-59)
and aims to quantify the extent to which it is important to take the dyadic nature of sleep-
wake cycles into account. It achieves this through two interrelated aims: (i) to use
Hierarchical Linear Models to measure dyadic interdependence in actigraphically recorded
variables; and (ii) to investigate how much of this dyadic interdependence truly results from
couple dynamics. The variables with the most significant couple interdependency were
‘Actual bed time’, ‘Sleep latency’, ‘Light/Dark ratio’ and ‘Wake bouts’. The paper
concludes by suggesting that interdependence may be the defining feature of couples’ sleep
and that we need to employ analytic approaches which both acknowledge this and which are
sensitive to the possibilities that not all aspects of sleep will behave in the same way.
Key words: couples, sleep, actigraphy
A paradox exists within sleep research. On the one hand, it has long been recognised that
‘the common practice of sharing a bed might be a significant contributing factor to daytime
sleepiness and chronic sleep loss’ (Kushida 2004, p. 133). Nearly forty years ago Monroe
(1969) highlighted how sleep architecture improves when bed partners sleep alone. In his
study of 14 couples, Monroe found that lone sleeping significantly increased the amount of
stage 4 sleep and reduced the number of awakenings by 60 per cent.
Yet, despite the existence of a nascent literature on couples’ sleep quality (Cartwright and
Knight 1987; Pankhurst and Horne 1994; Beninati et al 1999; Ulfberg et al 2000; Parish and
Lyng 2003; Strawbridge et al 2004; Rosenblatt 2006; Dittami et al 2007; Venn 2007),
discussions of couples’ sleep-wake remain largely confined to review articles (Troxel et al
2007), ‘brainstorms’ (Adams and Cromwell 1978, p. 11) or studies of marital quality (Larson
et al 1991). It is still the case that ‘almost everything that has been published in the social and
behavioural sciences and in medicine about adult sleep has looked at adult sleep as an
individual phenomenon’ (Rosenblatt 2006, p. 1), with studies largely ignoring the potential
dyadic nature of bed partner’s sleep.
Given this contradiction, this paper aims to quantify the extent to which it is important to take
the dyadic nature of sleep-wake cycles into account. It achieves this through two interrelated
aims: (i) to measure dyadic interdependence in actigraphically recorded variables; and (ii) to
investigate how much of this dyadic interdependence truly results from couple dynamics.
Forty heterosexual couples were recruited in South East England in 2004-2005. Flyers were
hand delivered around housing estates in the South East of England (eliciting an average
response of 1 couple per 100 flyers). Participants were also invited to pass on details of the
study to family and friends. Each participating couple received £100 to compensate them for
their time and commitment.
Inclusion criteria were that the male in the couple was aged between 20 and 59 inclusive and
must be married or cohabiting. Female subjects could also not knowingly be pregnant at the
beginning of the project. All participants continued their normal sleep/wake routine
throughout a one week data collection period and no restrictions were placed on activities,
food or drink. The study was approved by the University of Surrey ethics committee and
conformed to international ethical standards (Touitou et al 2006).
Due to non-compliance or missing data the analysis set included 36 couples. The age range
for the men was 21 years to 59 years (mean= 39.83; sem=1.92). The women ranged from 22
years to 58 years (mean=38.64; sem=1.86). Most partners were similar in age; the age
difference ranged from 0 years to 12 years (mean difference=2.83 years; sem=0.41).
Nineteen couples had no children living at home. Seventeen couples had children living at
home; 11 of which had at least one child under the age of 10 co-residing. All of the men
were in fulltime paid employment, meaning that, for the large part, the actigraphy data
included 5 paid employment days and 2 ‘rest’ days. Six of the women described themselves
as a ‘housewife’ or ‘not in employment’, three were full-time students and the remainder
were employed in a wide range of full or part-time jobs. The mean difference in Horne and
Östberg score (Horne and Östberg 1976), a measure of diurnal preference, was 7.82
(s.d.=6.42). This suggests that individuals were not selecting partners with similar diurnal
As well as completing the Horne and Östberg questionnaire and the Pittsburgh Sleep Quality
Index questionnaire (Buysse et al 1989), each partner wore an actigraph for 1 week.
Actigraphically recorded measurement data can, with high reliability, be translated into
information about a person’s sleep (Webster et al 1982; Cole et al 1992; Sadeh et al 1994;
Sadeh et al 1995; Blood et al 1997). Variables such as ‘sleep latency’, ‘sleep efficiency’,
‘number of awakenings’ and ‘sleep fragmentation’ are now routinely derived from actigraphy
data (Knutson et al 2007).
Within the present study, actiwatches (Cambridge Neurotechnology Ltd, [CNT] Cambridge,
UK) were set to collect data at 1 minute epochs. All watches were calibrated using
standardized equipment from CNT prior to use. Actigraphy files were downloaded and
summarised using inbuilt algorithms within the manufacturer’s software ‘Sleep Analysis 5’
(Cambridge Neurotechnology Ltd).
Following Morgenthaler et al’s (2007) recommendation that careful attention be paid to start
and stop times, three steps were taken to ensure the accuracy of our analysis: first, as all of
the sleep variables are calculated in relation to ‘bed time’ and ‘get up time’, participants were
asked to press an ‘event button’ (positioned on the surface of the watch) when they were in
bed ready to go to sleep and when they awoke in the morning. This event press was recorded
within the actigraph’s memory. Second, two experienced actigraphy scorers independently
analysed the files. Results were then compared, and any discrepancies between the two led
to further investigation. Particular attention was paid to those files where the subject did not
comply fully with marker press instructions and the person analysing was required to
estimate ‘bed time’ and/or ‘get up time’ (n=87 nights out of a total of 473 nights). Third,
audio diaries were referred to throughout the analysis and especially in cases where the ‘bed
time’ or ‘get up time’ were ambiguous. Although Lockley et al (1999) have shown that
subjective written diaries and actigraphy are poorly correlated in relation to number and
duration of awakenings the audio diaries within this study were differentiated from earlier
research, especially because bed partners were able to provide extensive data about each
Audio diaries (Sony Dictaphones) were given to each partner with the instruction that, for the
following 7 days, they should record, after awakening in the morning, any information they
wished about their sleep or wake during the previous night. Written instructions to this effect
were provided, which asked them specifically to record bed times, any disturbances
throughout the night, and number and times of awakening. Subjects were also instructed that
they could make multiple entries and return to the audio diary on as many occasions as they
wished throughout the day. The completed audio sleep diary recordings were transcribed in
full. A detailed discussion of this method can be found in Hislop et al (2005).
Twelve variables were calculated for each bed partner. These variables measure aspects of
sleep timing, sleep quality and circadian rhythmicity. The sleep timing variables included
‘Preferred bed time’ (taken from questionnaire items which asked about the time that partners
would like to go to bed), ‘Actual bed time’ and ‘Get up time’ (both of which were calculated
from marker presses within the actigraphy) and ‘Difference between preferred and actual
bedtime’. Sleep quality variables were all calculated from the actigraphy data and included
‘Sleep latency’, ‘Actual sleep time’, ‘Sleep efficiency’, ‘Wake bouts’ (number of times the
subject woke during the night), ‘Light/Dark ratio’ (the difference between amount of daytime
movement and nighttime movement) and ‘Fragmentation index’. As an indication of
circadian rhythmicity the ‘interdaily stability’ and ‘intradaily variability’ were calculated.
The ‘interdaily stability’ (IS) ‘gives an indication of the strength of the coupling between the
rest-activity rhythm and Zeitgebers’ (Van Someren et al 1997, p. 957). The ‘intradaily
variability’ (IV) ‘gives an indication of the fragmentation of the rhythm’ (Van Someren et al
1997, p. 957). Both variables were created following the method described in Van Someren
et al (1997).
Statistical approaches to examining the interdependence of couple data
There is debate surrounding how the interdependence of couples’ sleep should be measured.
Standard correlation analysis techniques are often applied to data on couples (Kenny and
Cook 1999), in which sleep data is aggregated to give an average score for each individual
within the couple. Each spouses’ aggregated score would then be correlated with their
partner’s aggregated score. With heterosexual dyads these techniques quantify the extent to
which women who receive a high score on a variable, relative to other women, are matched
with men who receive a high score, relative to other men. However, this aggregation may
result in cross-level errors or level of analysis errors (Gonzalez and Griffin 1997) and ignore
potentially interesting autocorrelations between nights.
Researchers also disagree over the appropriateness of using Hierarchical Linear Modelling to
examine couples’ sleep. Some authors suggest that it is a useful method for identifying and
handling the interdependence of married/cohabiting couples’ data (DeLongis and Holtzman
2005, p. 5-6). While others, such as Laurenceau and Bolger (2005) and Kenny et al (2006),
criticise the use of Hierarchical Linear Modelling within research on couples. In their
discussion of multilevel models with binary outcomes, McMahon et al (2006) suggest that it
is often not feasible to estimate random effects for both the intercept and slope because there
are only two observations per cluster (i.e. per couple). Further to this, dyads in which both
members have identical responses do not contribute to the likelihood function.
In light of these debates, and the aims of the paper, analysis was undertaken in three stages:
First, we began by inspecting the validity of the data. The Kolmogorov-Smirnov (K-S) test
was applied to test for normality (using SPSS v141.) and reliability measures were obtained
(using HLM for windows, version 6.02a; Raudenbush and Bryk 2002). Reliability measures
indicate whether the sample mean is a reliable estimate of the true mean. Reliability scores
range from 0 to 1 and reliability is considered medium if greater than 0.5 or large if greater
than 0.7. Residuals were also plotted to identify the number of couples which differed
significantly from the average at the 5% level. Outliers were removed from the data set prior
to further analysis. As Langford and Lewis (1998) suggest, the word ‘outlier’ is used by
different writers in different senses. Here we use the term to describe a situation where: i)
significant difference between couples exists; and ii) this significant difference between
couples no longer exists when a single couple is removed from the model.
Second, we examined the shared variance between partners within couples using
unconditional hierarchical linear models (Raudenbush and Bryk 2002). For each dependent
variable, intraclass correlations (ICC) were calculated to give an indication of the proportion
of the total variation that is accounted for by higher level units. For the two level models
(which analyse IS, IV, ‘Preferred bedtime’) we calculated the proportion of the total variance
which occurs between individuals (Level 1) and between couples (Level 2). For the three
level models (which include ‘Sleep efficiency’, ‘Wake bouts’, ‘Fragmentation index’ and
‘Actual bed time’) we calculated the portion of the total variation which occurs between
nights (Level 1), between individuals (Level 2) and between couples (Level 3) (cf. Singer
Because the ICC indicates proportion of variance at the couple level, it can be high even
when the actual amount of variation is small. Therefore, tests were performed to calculate
whether the proportion of the variation at the couple level is statistically significant. Each
model was run in MLwiN (version 2.02; Rasbash et al 2005) with and without the couple
level. The likelihood ratio statistic was then obtained by subtracting the -2*loglikelihoods of
the two models and comparing the result to a chi-squared distribution on 1 degree of freedom
(Rasbash et al 2005).
Finally, we investigated how much of the variance component at the couple level truly results
from couple dynamics. Couples in this study tended to be of similar age and, as age can
affect sleep and rhythms, some of the couple ICC could be due to similar ages of partners and
not specifically related to dyadic interdependence, shared events or direct impacts2. Huang
et al (2002), for example, suggest that the IV of those aged 61 – 91 is significantly higher
than those aged 21-44 and Monk et al (2006) found that compared to younger adults older
adults have a highly regular lifestyle. Therefore, ‘age’ was added to the models and changes
in the proportion of variance at the couple level noted. ‘Presence of children’ was also added
and changes in couple level variance observed. Due to differing rates of compliance among
the sleep measures used, the total number of participants within the models discussed below
Data was checked for validity. ‘Get up time’, ‘Sleep efficiency’ and ‘Difference between
preferred and actual bed time’ were not normally distributed. This was not resolved by log
transformation and no further analysis was performed on these three variables.
Figure 1 shows the number of couples which differ significantly from the average for the
remaining nine sleep variables. As Figure 1 illustrates, with interdaily stability (IS) only one
couple differed significantly from the average at the 5% level. Similarly, with intradaily
variability (IV) only 1 couple were significantly different from the average at the 5% level.
When these two couples were removed from the analysis there were no significant
differences between couples and a large proportion of the couple level variation was
removed. Therefore these two couples were considered outliers and removed from further
analysis of interdaily stability and intradaily variability.
Insert Figure 1 about here
Figure 1 also suggests that there is little couple level variation for ‘Actual sleep time’ and
‘Fragmentation index’ (as d=0). The residual plots in Figure 1 do indicate the existence of
some couple level variation in ‘Preferred bed time’ (d=3), ‘Sleep latency’ (d=4), ‘Light/Dark
ratio’ (d=4), and a high degree of couple level variation in ‘Wake bouts’ (d=8) and ‘Actual
bed time’ (d=17). This was confirmed by the unconditional models: ‘Preferred bed time’
(couple level ICC=0.42; p<0.01), ‘Sleep latency’ (couple level ICC=0.25, p<0.001),
‘Light/Dark ratio’ (couple level ICC=0.28, p<0.001), ‘Wake bouts’ (couple level ICC=0.42,
p<0.001) and ‘Actual bed time’ (couple level ICC=0.52, p<0.001) all showed significant
clustering at the couple level.
However, as Table 1 illustrates, after including ‘age’ and ‘presence of children’ in the
models, only ‘Sleep latency’, ‘Light/Dark ratio’, ‘Wake bouts’ and ‘Actual bed time’,
remained showing significant variance at the couple level3. Further to this, the proportion of
variance at the couple level for ‘Sleep latency’ reduced from 0.25 to 0.18.
Insert Table 1 about here
This paper has examined actigraphic data collected from 36 couples for one week. It has
investigated the partitioning of variance in twelve sleep/wake variables: ‘Preferred bed time’,
‘Actual bed time’, ‘Get up time’, ‘Difference between preferred and actual bed time’, ‘Actual
sleep time; ‘Light/Dark ratio’, ‘Sleep latency’, ‘Sleep efficiency’, ‘Fragmentation index’,
‘Number of wake bouts’, ‘Intradaily variability’ and ‘Interdaily stability’. ‘Get up time’,
‘Sleep efficiency’ and ‘Difference between preferred and actual bed time’ were not normally
distributed and no further analysis was performed on these variables.
After adding covariates to the models, four of the nine remaining variables showed
significant clustering at the couple level: ‘Actual bed time’, ‘Sleep latency’, ‘Light/Dark
ratio’ and ‘Wake bouts’. These findings extend those of previous studies. Crossley (2004, p.
18) suggests that ‘In situations of co-presence the parties to that situation need to secure co-
operation from one another for their own sleep ritual, whether this means common bedtimes
and sleep conditions or different but complementary patterns, with each party respecting the
needs of the other.’ Within the present study a large proportion of the variance for ‘actual
bed time’ resided at the couple level; suggesting a couple dynamic in the timing of bed. Yet,
for ‘Preferred bed time’ the largest proportion of the variance was at the individual level,
suggesting that couples do not necessarily want to go to their shared bed at the same time.
The results for ‘Wake bouts’ confirm earlier research which suggests that about one third of
nocturnal awakenings are similar to bed partners. In two related studies, Pankhurst and Horne
(1994) examined actigraphically recorded concordance in movement in 46 pairs of bed
partners and the differences in nocturnal movement in those with and without bed partners.
The authors concluded that approximately one-third of measured movements were common
to both bed-partners, although couples did not seem aware of this concordance and reported
sleeping better when their partner was there.
Where this study differs is that it quantifies those variables which are susceptible to dyadic
influence, illustrating the strength of the clustering whilst isolating the potential effects of
potential cofounders, such as age. It illustrates how social factors impinge on the timing (and
quality) of sleep-wake cycles and that bed partners are an important part of these social
factors. Our findings exemplify how the dyadic nature of couples’ sleep needs to be given
much more attention. As Troxel et al (2007, p. 389) suggest, ‘recognizing the dyadic nature
of sleep and incorporating such knowledge into both clinical practice and research in sleep
medicine may elucidate key mechanisms in the etiology and maintenance of both sleep
disorders and relationship problems and may ultimately inform novel treatments’. Indeed, a
small scale study by Cartwright (2008) found that continuous positive airway pressure
(CPAP) compliance in married men is (positively) related to the frequency with which his
partner sleeps with him. Interdependence between partners within couples is not simply a
statistical nuisance or a logistical problem. Interdependence may be the defining feature of
relationships (Kenny and Cook 1999) and, in societies where it is common for adult partners
to share a bed (Hislop 2007), it is perhaps also the defining feature of sleep.
1 – Three different software packages were used for the analysis (SPSS, HLM and MLwiN).
HLM was used for the reliability measures, K-S test results were obtained from SPSS v14
and the variance component information was taken from MLwiN. No single package could
offer all the information required. For example, HLM is one of the few packages which
provides reliability measures, yet the p-values and confidence intervals produced by HLM
differ from those obtained by most other programs (Hox 2002, p. 42).
2- We are grateful to one of the reviewers who pointed out the importance of addressing this
3- MCMC analysis confirmed the substantial interdependence in bed partner’s actual bed
time, sleep latency, light/dark ratio and wake bouts. MCMC is a conservative estimation
method in Hierarchical Linear Modelling.
The authors would like to offer sincere thanks to Vicky Vaughan for assisting with the data
collection and Ian Brunton-Smith and the reviewers for useful pointers and probing
questions. This study was funded by Economic and Social Research Council grant RES-000-
Adams BN, Cromwell RE. (1978). Morning and night people in the family: a preliminary
statement. Fam. Coord. 27:5-13
Beninati W, Harris CD, Herold DL, Shepard JW, Jr. (1999). The effect of snoring and
obstructive sleep apnea on the sleep quality of bed partners. Mayo. Clin. Proc. 74:955-958
Blood ML, Sack RL, Percy DC, Pen JC. (1997). A comparison of sleep detection by wrist
actigraphy, behavioural response, and polysomnography. Sleep 6:388-395
Cartwright R. (2008). Sleeping together: a pilot study of the effects of shared sleeping on
adherence to CPAP treatment in obstructive sleep apnea. JCSM 4:123-127
Cartwright, RD, Knight, S. (1987). Silent partners: the wives of sleep apneic patients. Sleep
Buysse DJ, Reynolds CF, Monk TH, Buman SR, Kupfer DJ. (1989). Pittsburgh Sleep Quality
Index: a new instrument for psychiatric practice and research. Psychiat. Res. 28:193–213
Cole RJ, Kripke DF, Gruen W, Mullany DJ, Gillin JC. (1992). Automatic sleep/wake
identification from wrist activity. Sleep 15:461-9
Crossley N. (2004). Sleep, reflexive embodiment and social networks, Paper presented at the
first ESRC 'Sleep and Society' seminar, 3 December, University of Warwick
Delongis A, Holtzman S. (2005). Coping in context: the role of stress, social support, and
personality in coping. J. Pers. 73:1633-1656
Dittami J, Keckeis M, Machatschke I, Katina S, Zeitlhofer J, Kloesch G. (2007). Sex
differences in the reactions to sleeping in pairs versus sleeping alone in humans. Sleep Biol.
Gonzalez R, Griffin D. (1997). The statistics of interdependence: treating dyadic data with
respect. In SW, Duck (Ed.), Handbook of Personal Relationships: Theory, Research, and
Interventions (2nd Ed). Chichester: Wiley
Huang Y-L, Liu R-Y, Wang Q-S, Van Someren EJW, Xu H, Zhou J-N. (2002). Age-
associated difference in circadian sleep-wake and rest-activity rhythms. Physiol. Behav.
Hislop J. (2007). A bed of roses or a bed of thorns? Negotiating the couple relationship
through sleep. Sociol. Res. Online 12: http://www.socresonline.org.uk/12/5/2.html
Hislop J, Arber S, Meadows R, Venn S. (2005). Narratives of the night: the use of audio
diaries in researching sleep. Sociol. Res. Online
Horne JA, Östberg O. (1976). A self-assessment questionnaire to determine morning-
eveningness in human circadian rhythms. Int. J. Chronobiol. 4:97-110
Hox, J. (2002). Multilevel analysis: techniques and applications. Lawrence Erlbaum
Kenny DA, Cook W. (1999). Partner effects in relationship research: conceptual issues,
analytic difficulties, and illustrations. Pers. Relationship. 6:433-448.
Kenny DA, Kashy DA, Cook WL. (2006). Dyadic data analysis. London: Guilford Press
Knutson KL, Rathouz PJ, Lijing LY, Liu K, Lauderdale DS. (2007). Intra-individual daily
and yearly variability in actigraphically recorded sleep measures: the CARDIA Study. Sleep
Kushida CA. (2004). Sleep deprivation: clinical issues, pharmacology and sleep loss effects.
Informa Health Care
Langford IH, Lewis T. (1998). Outliers in multilevel data. J. Roy. Statistical Society 161:121-
Larson JH, Crane DR, Smith CW. (1991). Morning and night couples: the effect of wake and
sleep patterns on marital adjustment. J Marital Fam. Ther. 17:53-65.
Laurenceau J-P, Bolger N. (2005). Using diary methods to study marital and family
processes. J. Fam. Psychol. 19:86-97.
Lockley SW, Skene DJ, Arendt J. (1999). Comparison between subjective and actigraphic
measurement of sleep and sleep rhythms. J. Sleep Res. 8:175-183
McMahon JR, Pouget ER, Tortu S. (2006). A guide for multilevel modelling of dyadic data
with binary outcomes using SAS PROC MIXED. Comput. Stat. Data An.
Monk TH, Buysse DJ, Hall M, Nofzinger EA, Thompson WK, Mazumdar SA, Reynolds III
CF. (2006). Age-related differences in the lifestyle regularity of seniors experiencing
bereavement, care-giving, insomnia, and advancement into old-old age. Chronobiol. Int.
Monroe LJ. (1969). Transient changes in EEG sleep patterns of married good sleepers: the
effects of altering sleeping arrangement. Psychophysiology 6:330–7
Morgenthaler T, Aless C, Friedman L, Owens J, Kapur V, Boehlecke B, Brown T, Chesson
A, Coleman J, Lee-Chiong T, Pancer J, Swick TJ. (2007). Practice parameters for the use of
actigraphy in the assessment of sleep and sleep disorders: An update for 2007. Sleep 30:519-
Pankhurst FP, Horne JA. (1994). The influence of bed partners on movement during sleep.
Parish JM, Lyng PJ, (2003). Quality of life in bed partners of patients with obstructive sleep
apnea or hypopnea after treatment with continuous positive airway pressure. Chest 124:942-
Rasbash J, Steele F, Browne WJ, Prosser B. (2005). A users guide to MLwiN version 2.0.
University of Bristol
Raudenbush SW, Bryk AS. (2002). Hierarchical linear models: applications and data
analysis methods. (2nd ed). CA: Sage
Rosenblatt PC. (2006). Two in a bed: the social system of couple bed sharing. New York,
State University of New York Press
Sadeh A, Acebo C, Seifer R, Aytur S, Carskadon MA. (1995). Activity-based assessment of
sleep-wake patterns during the 1st year of life. Infant Behav. Dev. 18:329-337
Sadeh A, Sharkey KM, Carskadon MA. (1994). Activity based sleep-wake identification: an
empirical test of methodological issues. Sleep 17:201-207
Singer JD. (1998). Using SAS PROC MIXED to fit multilevel models, hierarchical models
and linear growth models. J Educ. Behav. Stat. 24:323-355
Strawbridge WJ, Shema SJ, Roberts RE. (2004). Impact of spouses sleep problems on
partners. Sleep 27:527-531
Troxel WM, Robles TF, Hall M, Buysse DJ. (2007). Marital quality and the marital bed:
examining the covariation between relationship quality and sleep. Sleep Med. Rev. 11:389-
Touitou Y, Smolensky MH, Portaluppi F. (2006). Ethics, standards and procedures in human
and animal research in chronobiology. Chronobiol. Int. 23:1083-1096.
Ulfberg J, Carter N, Talback M, Edling C. (2000). Adverse health effects among women
living with heavy snorers. Health Care Women Int. 21:81-90
Van Someren EJW, Kessler A, Mirmiran M, Swaab DF. (1997). Indirect bright light
improves circadian rest-activity rhythm disturbances in demented patients. Biol. Psychiat.
Venn S. (2007). ‘It’s ok for a man to snore.’ The influence of gender in addressing sleep
disruption in couples. Sociol. Res. Online 12 http://www.socresonline.org.uk/12/5/1.html
Webster JB, Kripke DF, Messi S, Mullany DJ, Wyborney G. (1982). An activity-based sleep
monitor system for ambulatory use. Sleep 5:389-99
Table 1: Hierarchical Linear Model Results for 36 couples (72 subjects) over one week –
after removing outliers and including ‘age’ and ‘presence of children’ as predictors.
This table illustrates the results from the Unconditional Hierarchical Linear Models. The descriptives include the number of individuals (or
nights for three level data) in the data set (N) and the mean, standard deviation (SD), min and max for each variable. The variance columns
provide the actual variance and the proportion of variance [ICC] at each level. Actual Bed Time, actual sleep time, sleep latency,
fragmentation index, Light/Dark ratio and wake bouts all include weekdays only. Sleep Latency and Light/Dark ratio were log transformed.
* indicates significant differences between couples. *p<0.5, **p<0.01, ***p<0.001
Figure 1 – Caterpillar plots for nine variables, showing each couple and whether they differ
significantly from the average at the 5% level.
The figures above plot the Couple Level residual (+/- 1.96 sd) against the ‘rank’. These figures are useful as they enable the identification
of the number of couples which differ significantly from the average at the 5% level (indicated on each graph by d=x). Where the error bars
do not cross 0 couples can be considered significantly different. Those above the 0 line have a significantly higher outcome on the
dependent variable and those below the 0 line have a significantly lower outcome on the dependent variable. There are fewer couples in the
‘preferred bed time’ graph, as residual plots do not include couples where the bed partners had exactly the same preferred bed time. That is,
because of the similarities in bed times across the data set as a whole, some couples are ranked exactly the same as other couples.